src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSw_VdwNone_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            None
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEwSw_VdwNone_GeomW4P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  90     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  91     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  95     real             *ewtab;
  96     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  97     real             rswitch_scalar,d_scalar;
  98     _fjsp_v2r8       itab_tmp;
  99     _fjsp_v2r8       dummy_mask,cutoff_mask;
 100     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 101     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 102     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 103
 104     x                = xx[0];
 105     f                = ff[0];
 106
 107     nri              = nlist->nri;
 108     iinr             = nlist->iinr;
 109     jindex           = nlist->jindex;
 110     jjnr             = nlist->jjnr;
 111     shiftidx         = nlist->shift;
 112     gid              = nlist->gid;
 113     shiftvec         = fr->shift_vec[0];
 114     fshift           = fr->fshift[0];
 115     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 116     charge           = mdatoms->chargeA;
 117
 118     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 119     ewtab            = fr->ic->tabq_coul_FDV0;
 120     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 121     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 122
 123     /* Setup water-specific parameters */
 124     inr              = nlist->iinr[0];
 125     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 126     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 127     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 128
 129     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 130     rcutoff_scalar   = fr->rcoulomb;
 131     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 132     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 133
 134     rswitch_scalar   = fr->rcoulomb_switch;
 135     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 136     /* Setup switch parameters */
 137     d_scalar         = rcutoff_scalar-rswitch_scalar;
 138     d                = gmx_fjsp_set1_v2r8(d_scalar);
 139     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 140     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 141     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 142     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 143     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 144     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 145
 146     /* Avoid stupid compiler warnings */
 147     jnrA = jnrB = 0;
 148     j_coord_offsetA = 0;
 149     j_coord_offsetB = 0;
 150
 151     outeriter        = 0;
 152     inneriter        = 0;
 153
 154     /* Start outer loop over neighborlists */
 155     for(iidx=0; iidx<nri; iidx++)
 156     {
 157         /* Load shift vector for this list */
 158         i_shift_offset   = DIM*shiftidx[iidx];
 159
 160         /* Load limits for loop over neighbors */
 161         j_index_start    = jindex[iidx];
 162         j_index_end      = jindex[iidx+1];
 163
 164         /* Get outer coordinate index */
 165         inr              = iinr[iidx];
 166         i_coord_offset   = DIM*inr;
 167
 168         /* Load i particle coords and add shift vector */
 169         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 170                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 171
 172         fix1             = _fjsp_setzero_v2r8();
 173         fiy1             = _fjsp_setzero_v2r8();
 174         fiz1             = _fjsp_setzero_v2r8();
 175         fix2             = _fjsp_setzero_v2r8();
 176         fiy2             = _fjsp_setzero_v2r8();
 177         fiz2             = _fjsp_setzero_v2r8();
 178         fix3             = _fjsp_setzero_v2r8();
 179         fiy3             = _fjsp_setzero_v2r8();
 180         fiz3             = _fjsp_setzero_v2r8();
 181
 182         /* Reset potential sums */
 183         velecsum         = _fjsp_setzero_v2r8();
 184
 185         /* Start inner kernel loop */
 186         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 187         {
 188
 189             /* Get j neighbor index, and coordinate index */
 190             jnrA             = jjnr[jidx];
 191             jnrB             = jjnr[jidx+1];
 192             j_coord_offsetA  = DIM*jnrA;
 193             j_coord_offsetB  = DIM*jnrB;
 194
 195             /* load j atom coordinates */
 196             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 197                                               &jx0,&jy0,&jz0);
 198
 199             /* Calculate displacement vector */
 200             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 201             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 202             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 203             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 204             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 205             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 206             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 207             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 208             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 209
 210             /* Calculate squared distance and things based on it */
 211             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 212             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 213             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 214
 215             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 216             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 217             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 218
 219             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 220             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 221             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 222
 223             /* Load parameters for j particles */
 224             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 225
 226             fjx0             = _fjsp_setzero_v2r8();
 227             fjy0             = _fjsp_setzero_v2r8();
 228             fjz0             = _fjsp_setzero_v2r8();
 229
 230             /**************************
 231              * CALCULATE INTERACTIONS *
 232              **************************/
 233
 234             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 235             {
 236
 237             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 238
 239             /* Compute parameters for interactions between i and j atoms */
 240             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 241
 242             /* EWALD ELECTROSTATICS */
 243
 244             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 245             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 246             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 247             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 248             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 249
 250             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 251             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 252             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 253             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 254             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 255             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 256             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 257             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 258             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 259             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 260
 261             d                = _fjsp_sub_v2r8(r10,rswitch);
 262             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 263             d2               = _fjsp_mul_v2r8(d,d);
 264             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 265
 266             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 267
 268             /* Evaluate switch function */
 269             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 270             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 271             velec            = _fjsp_mul_v2r8(velec,sw);
 272             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 273
 274             /* Update potential sum for this i atom from the interaction with this j atom. */
 275             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 276             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 277
 278             fscal            = felec;
 279
 280             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 281
 282             /* Update vectorial force */
 283             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 284             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 285             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 286
 287             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 288             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 289             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 290
 291             }
 292
 293             /**************************
 294              * CALCULATE INTERACTIONS *
 295              **************************/
 296
 297             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 298             {
 299
 300             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 301
 302             /* Compute parameters for interactions between i and j atoms */
 303             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 304
 305             /* EWALD ELECTROSTATICS */
 306
 307             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 308             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 309             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 310             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 311             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 312
 313             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 314             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 315             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 316             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 317             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 318             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 319             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 320             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 321             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 322             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 323
 324             d                = _fjsp_sub_v2r8(r20,rswitch);
 325             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 326             d2               = _fjsp_mul_v2r8(d,d);
 327             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 328
 329             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 330
 331             /* Evaluate switch function */
 332             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 333             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 334             velec            = _fjsp_mul_v2r8(velec,sw);
 335             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 336
 337             /* Update potential sum for this i atom from the interaction with this j atom. */
 338             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 339             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 340
 341             fscal            = felec;
 342
 343             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 344
 345             /* Update vectorial force */
 346             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 347             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 348             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 349
 350             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 351             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 352             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 353
 354             }
 355
 356             /**************************
 357              * CALCULATE INTERACTIONS *
 358              **************************/
 359
 360             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 361             {
 362
 363             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 364
 365             /* Compute parameters for interactions between i and j atoms */
 366             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 367
 368             /* EWALD ELECTROSTATICS */
 369
 370             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 371             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 372             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 373             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 374             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 375
 376             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 377             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 378             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 379             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 380             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 381             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 382             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 383             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 384             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 385             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 386
 387             d                = _fjsp_sub_v2r8(r30,rswitch);
 388             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 389             d2               = _fjsp_mul_v2r8(d,d);
 390             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 391
 392             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 393
 394             /* Evaluate switch function */
 395             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 396             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
 397             velec            = _fjsp_mul_v2r8(velec,sw);
 398             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 399
 400             /* Update potential sum for this i atom from the interaction with this j atom. */
 401             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 402             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 403
 404             fscal            = felec;
 405
 406             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 407
 408             /* Update vectorial force */
 409             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 410             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 411             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 412
 413             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 414             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 415             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 416
 417             }
 418
 419             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 420
 421             /* Inner loop uses 207 flops */
 422         }
 423
 424         if(jidx<j_index_end)
 425         {
 426
 427             jnrA             = jjnr[jidx];
 428             j_coord_offsetA  = DIM*jnrA;
 429
 430             /* load j atom coordinates */
 431             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 432                                               &jx0,&jy0,&jz0);
 433
 434             /* Calculate displacement vector */
 435             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 436             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 437             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 438             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 439             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 440             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 441             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 442             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 443             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 444
 445             /* Calculate squared distance and things based on it */
 446             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 447             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 448             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 449
 450             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 451             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 452             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 453
 454             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 455             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 456             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 457
 458             /* Load parameters for j particles */
 459             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 460
 461             fjx0             = _fjsp_setzero_v2r8();
 462             fjy0             = _fjsp_setzero_v2r8();
 463             fjz0             = _fjsp_setzero_v2r8();
 464
 465             /**************************
 466              * CALCULATE INTERACTIONS *
 467              **************************/
 468
 469             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 470             {
 471
 472             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 473
 474             /* Compute parameters for interactions between i and j atoms */
 475             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 476
 477             /* EWALD ELECTROSTATICS */
 478
 479             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 480             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 481             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 482             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 483             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 484
 485             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 486             ewtabD           = _fjsp_setzero_v2r8();
 487             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 488             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 489             ewtabFn          = _fjsp_setzero_v2r8();
 490             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 491             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 492             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 493             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 494             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 495
 496             d                = _fjsp_sub_v2r8(r10,rswitch);
 497             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 498             d2               = _fjsp_mul_v2r8(d,d);
 499             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 500
 501             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 502
 503             /* Evaluate switch function */
 504             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 505             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 506             velec            = _fjsp_mul_v2r8(velec,sw);
 507             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 508
 509             /* Update potential sum for this i atom from the interaction with this j atom. */
 510             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 511             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 512             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 513
 514             fscal            = felec;
 515
 516             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 517
 518             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 519
 520             /* Update vectorial force */
 521             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 522             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 523             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 524
 525             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 526             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 527             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 528
 529             }
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 536             {
 537
 538             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 539
 540             /* Compute parameters for interactions between i and j atoms */
 541             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 542
 543             /* EWALD ELECTROSTATICS */
 544
 545             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 546             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 547             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 548             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 549             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 550
 551             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 552             ewtabD           = _fjsp_setzero_v2r8();
 553             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 554             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 555             ewtabFn          = _fjsp_setzero_v2r8();
 556             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 557             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 558             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 559             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 560             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 561
 562             d                = _fjsp_sub_v2r8(r20,rswitch);
 563             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 564             d2               = _fjsp_mul_v2r8(d,d);
 565             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 566
 567             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 568
 569             /* Evaluate switch function */
 570             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 571             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 572             velec            = _fjsp_mul_v2r8(velec,sw);
 573             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 574
 575             /* Update potential sum for this i atom from the interaction with this j atom. */
 576             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 577             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 578             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 579
 580             fscal            = felec;
 581
 582             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 583
 584             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 585
 586             /* Update vectorial force */
 587             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 588             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 589             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 590
 591             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 592             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 593             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 594
 595             }
 596
 597             /**************************
 598              * CALCULATE INTERACTIONS *
 599              **************************/
 600
 601             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 602             {
 603
 604             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 605
 606             /* Compute parameters for interactions between i and j atoms */
 607             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 608
 609             /* EWALD ELECTROSTATICS */
 610
 611             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 612             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 613             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 614             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 615             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 616
 617             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 618             ewtabD           = _fjsp_setzero_v2r8();
 619             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 620             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 621             ewtabFn          = _fjsp_setzero_v2r8();
 622             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 623             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 624             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 625             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 626             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 627
 628             d                = _fjsp_sub_v2r8(r30,rswitch);
 629             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 630             d2               = _fjsp_mul_v2r8(d,d);
 631             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 632
 633             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 634
 635             /* Evaluate switch function */
 636             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 637             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
 638             velec            = _fjsp_mul_v2r8(velec,sw);
 639             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 640
 641             /* Update potential sum for this i atom from the interaction with this j atom. */
 642             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 643             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 644             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 645
 646             fscal            = felec;
 647
 648             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 649
 650             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 651
 652             /* Update vectorial force */
 653             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 654             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 655             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 656
 657             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 658             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 659             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 660
 661             }
 662
 663             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 664
 665             /* Inner loop uses 207 flops */
 666         }
 667
 668         /* End of innermost loop */
 669
 670         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 671                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
 672
 673         ggid                        = gid[iidx];
 674         /* Update potential energies */
 675         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 676
 677         /* Increment number of inner iterations */
 678         inneriter                  += j_index_end - j_index_start;
 679
 680         /* Outer loop uses 19 flops */
 681     }
 682
 683     /* Increment number of outer iterations */
 684     outeriter        += nri;
 685
 686     /* Update outer/inner flops */
 687
 688     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_VF,outeriter*19 + inneriter*207);
 689 }
 690 /*
 691  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 692  * Electrostatics interaction: Ewald
 693  * VdW interaction:            None
 694  * Geometry:                   Water4-Particle
 695  * Calculate force/pot:        Force
 696  */
 697 void
 698 nb_kernel_ElecEwSw_VdwNone_GeomW4P1_F_sparc64_hpc_ace_double
 699                     (t_nblist                    * gmx_restrict       nlist,
 700                      rvec                        * gmx_restrict          xx,
 701                      rvec                        * gmx_restrict          ff,
 702                      t_forcerec                  * gmx_restrict          fr,
 703                      t_mdatoms                   * gmx_restrict     mdatoms,
 704                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 705                      t_nrnb                      * gmx_restrict        nrnb)
 706 {
 707     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 708      * just 0 for non-waters.
 709      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 710      * jnr indices corresponding to data put in the four positions in the SIMD register.
 711      */
 712     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 713     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 714     int              jnrA,jnrB;
 715     int              j_coord_offsetA,j_coord_offsetB;
 716     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 717     real             rcutoff_scalar;
 718     real             *shiftvec,*fshift,*x,*f;
 719     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 720     int              vdwioffset1;
 721     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 722     int              vdwioffset2;
 723     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 724     int              vdwioffset3;
 725     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 726     int              vdwjidx0A,vdwjidx0B;
 727     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 728     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 729     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 730     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 731     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 732     real             *charge;
 733     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 734     real             *ewtab;
 735     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 736     real             rswitch_scalar,d_scalar;
 737     _fjsp_v2r8       itab_tmp;
 738     _fjsp_v2r8       dummy_mask,cutoff_mask;
 739     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 740     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 741     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 742
 743     x                = xx[0];
 744     f                = ff[0];
 745
 746     nri              = nlist->nri;
 747     iinr             = nlist->iinr;
 748     jindex           = nlist->jindex;
 749     jjnr             = nlist->jjnr;
 750     shiftidx         = nlist->shift;
 751     gid              = nlist->gid;
 752     shiftvec         = fr->shift_vec[0];
 753     fshift           = fr->fshift[0];
 754     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 755     charge           = mdatoms->chargeA;
 756
 757     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 758     ewtab            = fr->ic->tabq_coul_FDV0;
 759     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 760     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 761
 762     /* Setup water-specific parameters */
 763     inr              = nlist->iinr[0];
 764     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 765     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 766     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 767
 768     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 769     rcutoff_scalar   = fr->rcoulomb;
 770     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 771     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 772
 773     rswitch_scalar   = fr->rcoulomb_switch;
 774     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 775     /* Setup switch parameters */
 776     d_scalar         = rcutoff_scalar-rswitch_scalar;
 777     d                = gmx_fjsp_set1_v2r8(d_scalar);
 778     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 779     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 780     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 781     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 782     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 783     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 784
 785     /* Avoid stupid compiler warnings */
 786     jnrA = jnrB = 0;
 787     j_coord_offsetA = 0;
 788     j_coord_offsetB = 0;
 789
 790     outeriter        = 0;
 791     inneriter        = 0;
 792
 793     /* Start outer loop over neighborlists */
 794     for(iidx=0; iidx<nri; iidx++)
 795     {
 796         /* Load shift vector for this list */
 797         i_shift_offset   = DIM*shiftidx[iidx];
 798
 799         /* Load limits for loop over neighbors */
 800         j_index_start    = jindex[iidx];
 801         j_index_end      = jindex[iidx+1];
 802
 803         /* Get outer coordinate index */
 804         inr              = iinr[iidx];
 805         i_coord_offset   = DIM*inr;
 806
 807         /* Load i particle coords and add shift vector */
 808         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset+DIM,
 809                                                  &ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 810
 811         fix1             = _fjsp_setzero_v2r8();
 812         fiy1             = _fjsp_setzero_v2r8();
 813         fiz1             = _fjsp_setzero_v2r8();
 814         fix2             = _fjsp_setzero_v2r8();
 815         fiy2             = _fjsp_setzero_v2r8();
 816         fiz2             = _fjsp_setzero_v2r8();
 817         fix3             = _fjsp_setzero_v2r8();
 818         fiy3             = _fjsp_setzero_v2r8();
 819         fiz3             = _fjsp_setzero_v2r8();
 820
 821         /* Start inner kernel loop */
 822         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 823         {
 824
 825             /* Get j neighbor index, and coordinate index */
 826             jnrA             = jjnr[jidx];
 827             jnrB             = jjnr[jidx+1];
 828             j_coord_offsetA  = DIM*jnrA;
 829             j_coord_offsetB  = DIM*jnrB;
 830
 831             /* load j atom coordinates */
 832             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 833                                               &jx0,&jy0,&jz0);
 834
 835             /* Calculate displacement vector */
 836             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 837             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 838             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 839             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 840             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 841             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 842             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 843             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 844             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 845
 846             /* Calculate squared distance and things based on it */
 847             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 848             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 849             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 850
 851             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 852             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 853             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 854
 855             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 856             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 857             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 858
 859             /* Load parameters for j particles */
 860             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 861
 862             fjx0             = _fjsp_setzero_v2r8();
 863             fjy0             = _fjsp_setzero_v2r8();
 864             fjz0             = _fjsp_setzero_v2r8();
 865
 866             /**************************
 867              * CALCULATE INTERACTIONS *
 868              **************************/
 869
 870             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 871             {
 872
 873             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 874
 875             /* Compute parameters for interactions between i and j atoms */
 876             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 877
 878             /* EWALD ELECTROSTATICS */
 879
 880             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 881             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 882             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 883             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 884             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 885
 886             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 887             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 888             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 889             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 890             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 891             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 892             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 893             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 894             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 895             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 896
 897             d                = _fjsp_sub_v2r8(r10,rswitch);
 898             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 899             d2               = _fjsp_mul_v2r8(d,d);
 900             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 901
 902             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 903
 904             /* Evaluate switch function */
 905             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 906             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 907             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 908
 909             fscal            = felec;
 910
 911             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 912
 913             /* Update vectorial force */
 914             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 915             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 916             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 917
 918             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 919             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 920             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 921
 922             }
 923
 924             /**************************
 925              * CALCULATE INTERACTIONS *
 926              **************************/
 927
 928             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 929             {
 930
 931             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 932
 933             /* Compute parameters for interactions between i and j atoms */
 934             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 935
 936             /* EWALD ELECTROSTATICS */
 937
 938             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 939             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 940             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 941             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 942             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 943
 944             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 945             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 946             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 947             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 948             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 949             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 950             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 951             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 952             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 953             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 954
 955             d                = _fjsp_sub_v2r8(r20,rswitch);
 956             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 957             d2               = _fjsp_mul_v2r8(d,d);
 958             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 959
 960             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 961
 962             /* Evaluate switch function */
 963             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 964             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 965             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 966
 967             fscal            = felec;
 968
 969             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 970
 971             /* Update vectorial force */
 972             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 973             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 974             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 975
 976             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 977             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 978             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 979
 980             }
 981
 982             /**************************
 983              * CALCULATE INTERACTIONS *
 984              **************************/
 985
 986             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 987             {
 988
 989             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 990
 991             /* Compute parameters for interactions between i and j atoms */
 992             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 993
 994             /* EWALD ELECTROSTATICS */
 995
 996             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 997             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 998             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 999             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1000             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1001
1002             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1003             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1004             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1005             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1006             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1007             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1008             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1009             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1010             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
1011             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1012
1013             d                = _fjsp_sub_v2r8(r30,rswitch);
1014             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1015             d2               = _fjsp_mul_v2r8(d,d);
1016             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1017
1018             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1019
1020             /* Evaluate switch function */
1021             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1022             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
1023             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1024
1025             fscal            = felec;
1026
1027             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1028
1029             /* Update vectorial force */
1030             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1031             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1032             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1033
1034             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1035             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1036             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1037
1038             }
1039
1040             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1041
1042             /* Inner loop uses 198 flops */
1043         }
1044
1045         if(jidx<j_index_end)
1046         {
1047
1048             jnrA             = jjnr[jidx];
1049             j_coord_offsetA  = DIM*jnrA;
1050
1051             /* load j atom coordinates */
1052             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1053                                               &jx0,&jy0,&jz0);
1054
1055             /* Calculate displacement vector */
1056             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1057             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1058             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1059             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1060             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1061             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1062             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1063             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1064             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1065
1066             /* Calculate squared distance and things based on it */
1067             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1068             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1069             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1070
1071             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1072             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1073             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1074
1075             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1076             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1077             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1078
1079             /* Load parameters for j particles */
1080             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1081
1082             fjx0             = _fjsp_setzero_v2r8();
1083             fjy0             = _fjsp_setzero_v2r8();
1084             fjz0             = _fjsp_setzero_v2r8();
1085
1086             /**************************
1087              * CALCULATE INTERACTIONS *
1088              **************************/
1089
1090             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1091             {
1092
1093             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1094
1095             /* Compute parameters for interactions between i and j atoms */
1096             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1097
1098             /* EWALD ELECTROSTATICS */
1099
1100             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1101             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1102             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1103             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1104             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1105
1106             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1107             ewtabD           = _fjsp_setzero_v2r8();
1108             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1109             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1110             ewtabFn          = _fjsp_setzero_v2r8();
1111             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1112             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1113             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1114             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1115             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1116
1117             d                = _fjsp_sub_v2r8(r10,rswitch);
1118             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1119             d2               = _fjsp_mul_v2r8(d,d);
1120             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1121
1122             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1123
1124             /* Evaluate switch function */
1125             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1126             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1127             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1128
1129             fscal            = felec;
1130
1131             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1132
1133             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1134
1135             /* Update vectorial force */
1136             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1137             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1138             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1139
1140             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1141             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1142             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1143
1144             }
1145
1146             /**************************
1147              * CALCULATE INTERACTIONS *
1148              **************************/
1149
1150             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1151             {
1152
1153             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1154
1155             /* Compute parameters for interactions between i and j atoms */
1156             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1157
1158             /* EWALD ELECTROSTATICS */
1159
1160             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1161             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1162             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1163             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1164             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1165
1166             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1167             ewtabD           = _fjsp_setzero_v2r8();
1168             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1169             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1170             ewtabFn          = _fjsp_setzero_v2r8();
1171             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1172             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1173             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1174             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1175             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1176
1177             d                = _fjsp_sub_v2r8(r20,rswitch);
1178             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1179             d2               = _fjsp_mul_v2r8(d,d);
1180             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1181
1182             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1183
1184             /* Evaluate switch function */
1185             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1186             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1187             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1188
1189             fscal            = felec;
1190
1191             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1192
1193             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1194
1195             /* Update vectorial force */
1196             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1197             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1198             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1199
1200             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1201             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1202             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1203
1204             }
1205
1206             /**************************
1207              * CALCULATE INTERACTIONS *
1208              **************************/
1209
1210             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1211             {
1212
1213             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1214
1215             /* Compute parameters for interactions between i and j atoms */
1216             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1217
1218             /* EWALD ELECTROSTATICS */
1219
1220             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1221             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1222             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1223             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1224             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1225
1226             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1227             ewtabD           = _fjsp_setzero_v2r8();
1228             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1229             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1230             ewtabFn          = _fjsp_setzero_v2r8();
1231             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1232             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1233             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1234             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
1235             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1236
1237             d                = _fjsp_sub_v2r8(r30,rswitch);
1238             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1239             d2               = _fjsp_mul_v2r8(d,d);
1240             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1241
1242             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1243
1244             /* Evaluate switch function */
1245             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1246             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv30,_fjsp_mul_v2r8(velec,dsw)) );
1247             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1248
1249             fscal            = felec;
1250
1251             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1252
1253             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1254
1255             /* Update vectorial force */
1256             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1257             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1258             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1259
1260             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1261             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1262             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1263
1264             }
1265
1266             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1267
1268             /* Inner loop uses 198 flops */
1269         }
1270
1271         /* End of innermost loop */
1272
1273         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1274                                               f+i_coord_offset+DIM,fshift+i_shift_offset);
1275
1276         /* Increment number of inner iterations */
1277         inneriter                  += j_index_end - j_index_start;
1278
1279         /* Outer loop uses 18 flops */
1280     }
1281
1282     /* Increment number of outer iterations */
1283     outeriter        += nri;
1284
1285     /* Update outer/inner flops */
1286
1287     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W4_F,outeriter*18 + inneriter*198);
1288 }