src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSw_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_ElecEwSw_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_ElecEwSw_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       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  95     real             rswitch_scalar,d_scalar;
  96     _fjsp_v2r8       itab_tmp;
  97     _fjsp_v2r8       dummy_mask,cutoff_mask;
  98     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  99     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 100     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 101
 102     x                = xx[0];
 103     f                = ff[0];
 104
 105     nri              = nlist->nri;
 106     iinr             = nlist->iinr;
 107     jindex           = nlist->jindex;
 108     jjnr             = nlist->jjnr;
 109     shiftidx         = nlist->shift;
 110     gid              = nlist->gid;
 111     shiftvec         = fr->shift_vec[0];
 112     fshift           = fr->fshift[0];
 113     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 114     charge           = mdatoms->chargeA;
 115
 116     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 117     ewtab            = fr->ic->tabq_coul_FDV0;
 118     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 119     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 120
 121     /* Setup water-specific parameters */
 122     inr              = nlist->iinr[0];
 123     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 124     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 125     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 126
 127     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 128     rcutoff_scalar   = fr->rcoulomb;
 129     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 130     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 131
 132     rswitch_scalar   = fr->rcoulomb_switch;
 133     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 134     /* Setup switch parameters */
 135     d_scalar         = rcutoff_scalar-rswitch_scalar;
 136     d                = gmx_fjsp_set1_v2r8(d_scalar);
 137     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 138     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 139     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 140     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 141     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 142     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 143
 144     /* Avoid stupid compiler warnings */
 145     jnrA = jnrB = 0;
 146     j_coord_offsetA = 0;
 147     j_coord_offsetB = 0;
 148
 149     outeriter        = 0;
 150     inneriter        = 0;
 151
 152     /* Start outer loop over neighborlists */
 153     for(iidx=0; iidx<nri; iidx++)
 154     {
 155         /* Load shift vector for this list */
 156         i_shift_offset   = DIM*shiftidx[iidx];
 157
 158         /* Load limits for loop over neighbors */
 159         j_index_start    = jindex[iidx];
 160         j_index_end      = jindex[iidx+1];
 161
 162         /* Get outer coordinate index */
 163         inr              = iinr[iidx];
 164         i_coord_offset   = DIM*inr;
 165
 166         /* Load i particle coords and add shift vector */
 167         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 168                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 169
 170         fix0             = _fjsp_setzero_v2r8();
 171         fiy0             = _fjsp_setzero_v2r8();
 172         fiz0             = _fjsp_setzero_v2r8();
 173         fix1             = _fjsp_setzero_v2r8();
 174         fiy1             = _fjsp_setzero_v2r8();
 175         fiz1             = _fjsp_setzero_v2r8();
 176         fix2             = _fjsp_setzero_v2r8();
 177         fiy2             = _fjsp_setzero_v2r8();
 178         fiz2             = _fjsp_setzero_v2r8();
 179
 180         /* Reset potential sums */
 181         velecsum         = _fjsp_setzero_v2r8();
 182
 183         /* Start inner kernel loop */
 184         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 185         {
 186
 187             /* Get j neighbor index, and coordinate index */
 188             jnrA             = jjnr[jidx];
 189             jnrB             = jjnr[jidx+1];
 190             j_coord_offsetA  = DIM*jnrA;
 191             j_coord_offsetB  = DIM*jnrB;
 192
 193             /* load j atom coordinates */
 194             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 195                                               &jx0,&jy0,&jz0);
 196
 197             /* Calculate displacement vector */
 198             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 199             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 200             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 201             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 202             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 203             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 204             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 205             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 206             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 207
 208             /* Calculate squared distance and things based on it */
 209             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 210             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 211             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 212
 213             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 214             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 215             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 216
 217             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 218             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 219             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 220
 221             /* Load parameters for j particles */
 222             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 223
 224             fjx0             = _fjsp_setzero_v2r8();
 225             fjy0             = _fjsp_setzero_v2r8();
 226             fjz0             = _fjsp_setzero_v2r8();
 227
 228             /**************************
 229              * CALCULATE INTERACTIONS *
 230              **************************/
 231
 232             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 233             {
 234
 235             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 236
 237             /* Compute parameters for interactions between i and j atoms */
 238             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 239
 240             /* EWALD ELECTROSTATICS */
 241
 242             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 243             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 244             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 245             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 246             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 247
 248             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 249             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 250             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 251             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 252             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 253             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 254             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 255             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 256             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 257             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 258
 259             d                = _fjsp_sub_v2r8(r00,rswitch);
 260             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 261             d2               = _fjsp_mul_v2r8(d,d);
 262             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 263
 264             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 265
 266             /* Evaluate switch function */
 267             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 268             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 269             velec            = _fjsp_mul_v2r8(velec,sw);
 270             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 271
 272             /* Update potential sum for this i atom from the interaction with this j atom. */
 273             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 274             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 275
 276             fscal            = felec;
 277
 278             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 279
 280             /* Update vectorial force */
 281             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 282             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 283             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 284
 285             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 286             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 287             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 288
 289             }
 290
 291             /**************************
 292              * CALCULATE INTERACTIONS *
 293              **************************/
 294
 295             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 296             {
 297
 298             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 299
 300             /* Compute parameters for interactions between i and j atoms */
 301             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 302
 303             /* EWALD ELECTROSTATICS */
 304
 305             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 306             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 307             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 308             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 309             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 310
 311             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 312             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 313             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 314             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 315             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 316             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 317             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 318             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 319             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 320             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 321
 322             d                = _fjsp_sub_v2r8(r10,rswitch);
 323             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 324             d2               = _fjsp_mul_v2r8(d,d);
 325             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 326
 327             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 328
 329             /* Evaluate switch function */
 330             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 331             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 332             velec            = _fjsp_mul_v2r8(velec,sw);
 333             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 334
 335             /* Update potential sum for this i atom from the interaction with this j atom. */
 336             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 337             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 338
 339             fscal            = felec;
 340
 341             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 342
 343             /* Update vectorial force */
 344             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 345             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 346             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 347
 348             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 349             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 350             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 351
 352             }
 353
 354             /**************************
 355              * CALCULATE INTERACTIONS *
 356              **************************/
 357
 358             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 359             {
 360
 361             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 362
 363             /* Compute parameters for interactions between i and j atoms */
 364             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 365
 366             /* EWALD ELECTROSTATICS */
 367
 368             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 369             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 370             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 371             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 372             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 373
 374             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 375             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 376             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 377             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 378             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 379             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 380             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 381             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 382             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 383             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 384
 385             d                = _fjsp_sub_v2r8(r20,rswitch);
 386             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 387             d2               = _fjsp_mul_v2r8(d,d);
 388             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 389
 390             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 391
 392             /* Evaluate switch function */
 393             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 394             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 395             velec            = _fjsp_mul_v2r8(velec,sw);
 396             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 397
 398             /* Update potential sum for this i atom from the interaction with this j atom. */
 399             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 400             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 401
 402             fscal            = felec;
 403
 404             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 405
 406             /* Update vectorial force */
 407             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 408             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 409             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 410
 411             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 412             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 413             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 414
 415             }
 416
 417             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 418
 419             /* Inner loop uses 207 flops */
 420         }
 421
 422         if(jidx<j_index_end)
 423         {
 424
 425             jnrA             = jjnr[jidx];
 426             j_coord_offsetA  = DIM*jnrA;
 427
 428             /* load j atom coordinates */
 429             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 430                                               &jx0,&jy0,&jz0);
 431
 432             /* Calculate displacement vector */
 433             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 434             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 435             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 436             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 437             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 438             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 439             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 440             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 441             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 442
 443             /* Calculate squared distance and things based on it */
 444             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 445             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 446             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 447
 448             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 449             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 450             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 451
 452             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 453             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 454             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 455
 456             /* Load parameters for j particles */
 457             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 458
 459             fjx0             = _fjsp_setzero_v2r8();
 460             fjy0             = _fjsp_setzero_v2r8();
 461             fjz0             = _fjsp_setzero_v2r8();
 462
 463             /**************************
 464              * CALCULATE INTERACTIONS *
 465              **************************/
 466
 467             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 468             {
 469
 470             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 471
 472             /* Compute parameters for interactions between i and j atoms */
 473             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 474
 475             /* EWALD ELECTROSTATICS */
 476
 477             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 478             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 479             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 480             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 481             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 482
 483             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 484             ewtabD           = _fjsp_setzero_v2r8();
 485             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 486             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 487             ewtabFn          = _fjsp_setzero_v2r8();
 488             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 489             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 490             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 491             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 492             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 493
 494             d                = _fjsp_sub_v2r8(r00,rswitch);
 495             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 496             d2               = _fjsp_mul_v2r8(d,d);
 497             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 498
 499             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 500
 501             /* Evaluate switch function */
 502             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 503             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 504             velec            = _fjsp_mul_v2r8(velec,sw);
 505             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 506
 507             /* Update potential sum for this i atom from the interaction with this j atom. */
 508             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 509             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 510             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 511
 512             fscal            = felec;
 513
 514             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 515
 516             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 517
 518             /* Update vectorial force */
 519             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 520             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 521             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 522
 523             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 524             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 525             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 526
 527             }
 528
 529             /**************************
 530              * CALCULATE INTERACTIONS *
 531              **************************/
 532
 533             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 534             {
 535
 536             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 537
 538             /* Compute parameters for interactions between i and j atoms */
 539             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 540
 541             /* EWALD ELECTROSTATICS */
 542
 543             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 544             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 545             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 546             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 547             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 548
 549             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 550             ewtabD           = _fjsp_setzero_v2r8();
 551             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 552             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 553             ewtabFn          = _fjsp_setzero_v2r8();
 554             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 555             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 556             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 557             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 558             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 559
 560             d                = _fjsp_sub_v2r8(r10,rswitch);
 561             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 562             d2               = _fjsp_mul_v2r8(d,d);
 563             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 564
 565             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 566
 567             /* Evaluate switch function */
 568             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 569             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 570             velec            = _fjsp_mul_v2r8(velec,sw);
 571             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 572
 573             /* Update potential sum for this i atom from the interaction with this j atom. */
 574             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 575             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 576             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 577
 578             fscal            = felec;
 579
 580             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 581
 582             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 583
 584             /* Update vectorial force */
 585             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 586             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 587             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 588
 589             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 590             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 591             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 592
 593             }
 594
 595             /**************************
 596              * CALCULATE INTERACTIONS *
 597              **************************/
 598
 599             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 600             {
 601
 602             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 603
 604             /* Compute parameters for interactions between i and j atoms */
 605             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 606
 607             /* EWALD ELECTROSTATICS */
 608
 609             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 610             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 611             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 612             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 613             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 614
 615             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 616             ewtabD           = _fjsp_setzero_v2r8();
 617             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 618             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 619             ewtabFn          = _fjsp_setzero_v2r8();
 620             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 621             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 622             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 623             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 624             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 625
 626             d                = _fjsp_sub_v2r8(r20,rswitch);
 627             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 628             d2               = _fjsp_mul_v2r8(d,d);
 629             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 630
 631             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 632
 633             /* Evaluate switch function */
 634             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 635             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
 636             velec            = _fjsp_mul_v2r8(velec,sw);
 637             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 638
 639             /* Update potential sum for this i atom from the interaction with this j atom. */
 640             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 641             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 642             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 643
 644             fscal            = felec;
 645
 646             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 647
 648             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 649
 650             /* Update vectorial force */
 651             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 652             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 653             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 654
 655             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 656             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 657             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 658
 659             }
 660
 661             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 662
 663             /* Inner loop uses 207 flops */
 664         }
 665
 666         /* End of innermost loop */
 667
 668         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 669                                               f+i_coord_offset,fshift+i_shift_offset);
 670
 671         ggid                        = gid[iidx];
 672         /* Update potential energies */
 673         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 674
 675         /* Increment number of inner iterations */
 676         inneriter                  += j_index_end - j_index_start;
 677
 678         /* Outer loop uses 19 flops */
 679     }
 680
 681     /* Increment number of outer iterations */
 682     outeriter        += nri;
 683
 684     /* Update outer/inner flops */
 685
 686     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_VF,outeriter*19 + inneriter*207);
 687 }
 688 /*
 689  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 690  * Electrostatics interaction: Ewald
 691  * VdW interaction:            None
 692  * Geometry:                   Water3-Particle
 693  * Calculate force/pot:        Force
 694  */
 695 void
 696 nb_kernel_ElecEwSw_VdwNone_GeomW3P1_F_sparc64_hpc_ace_double
 697                     (t_nblist                    * gmx_restrict       nlist,
 698                      rvec                        * gmx_restrict          xx,
 699                      rvec                        * gmx_restrict          ff,
 700                      t_forcerec                  * gmx_restrict          fr,
 701                      t_mdatoms                   * gmx_restrict     mdatoms,
 702                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 703                      t_nrnb                      * gmx_restrict        nrnb)
 704 {
 705     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 706      * just 0 for non-waters.
 707      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 708      * jnr indices corresponding to data put in the four positions in the SIMD register.
 709      */
 710     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 711     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 712     int              jnrA,jnrB;
 713     int              j_coord_offsetA,j_coord_offsetB;
 714     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 715     real             rcutoff_scalar;
 716     real             *shiftvec,*fshift,*x,*f;
 717     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 718     int              vdwioffset0;
 719     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 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              vdwjidx0A,vdwjidx0B;
 725     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 726     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 727     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 728     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 729     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 730     real             *charge;
 731     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 732     real             *ewtab;
 733     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 734     real             rswitch_scalar,d_scalar;
 735     _fjsp_v2r8       itab_tmp;
 736     _fjsp_v2r8       dummy_mask,cutoff_mask;
 737     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 738     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 739     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 740
 741     x                = xx[0];
 742     f                = ff[0];
 743
 744     nri              = nlist->nri;
 745     iinr             = nlist->iinr;
 746     jindex           = nlist->jindex;
 747     jjnr             = nlist->jjnr;
 748     shiftidx         = nlist->shift;
 749     gid              = nlist->gid;
 750     shiftvec         = fr->shift_vec[0];
 751     fshift           = fr->fshift[0];
 752     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 753     charge           = mdatoms->chargeA;
 754
 755     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 756     ewtab            = fr->ic->tabq_coul_FDV0;
 757     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 758     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 759
 760     /* Setup water-specific parameters */
 761     inr              = nlist->iinr[0];
 762     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 763     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 764     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 765
 766     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 767     rcutoff_scalar   = fr->rcoulomb;
 768     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 769     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 770
 771     rswitch_scalar   = fr->rcoulomb_switch;
 772     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 773     /* Setup switch parameters */
 774     d_scalar         = rcutoff_scalar-rswitch_scalar;
 775     d                = gmx_fjsp_set1_v2r8(d_scalar);
 776     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 777     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 778     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 779     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 780     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 781     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 782
 783     /* Avoid stupid compiler warnings */
 784     jnrA = jnrB = 0;
 785     j_coord_offsetA = 0;
 786     j_coord_offsetB = 0;
 787
 788     outeriter        = 0;
 789     inneriter        = 0;
 790
 791     /* Start outer loop over neighborlists */
 792     for(iidx=0; iidx<nri; iidx++)
 793     {
 794         /* Load shift vector for this list */
 795         i_shift_offset   = DIM*shiftidx[iidx];
 796
 797         /* Load limits for loop over neighbors */
 798         j_index_start    = jindex[iidx];
 799         j_index_end      = jindex[iidx+1];
 800
 801         /* Get outer coordinate index */
 802         inr              = iinr[iidx];
 803         i_coord_offset   = DIM*inr;
 804
 805         /* Load i particle coords and add shift vector */
 806         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 807                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 808
 809         fix0             = _fjsp_setzero_v2r8();
 810         fiy0             = _fjsp_setzero_v2r8();
 811         fiz0             = _fjsp_setzero_v2r8();
 812         fix1             = _fjsp_setzero_v2r8();
 813         fiy1             = _fjsp_setzero_v2r8();
 814         fiz1             = _fjsp_setzero_v2r8();
 815         fix2             = _fjsp_setzero_v2r8();
 816         fiy2             = _fjsp_setzero_v2r8();
 817         fiz2             = _fjsp_setzero_v2r8();
 818
 819         /* Start inner kernel loop */
 820         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 821         {
 822
 823             /* Get j neighbor index, and coordinate index */
 824             jnrA             = jjnr[jidx];
 825             jnrB             = jjnr[jidx+1];
 826             j_coord_offsetA  = DIM*jnrA;
 827             j_coord_offsetB  = DIM*jnrB;
 828
 829             /* load j atom coordinates */
 830             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 831                                               &jx0,&jy0,&jz0);
 832
 833             /* Calculate displacement vector */
 834             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 835             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 836             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 837             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 838             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 839             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 840             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 841             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 842             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 843
 844             /* Calculate squared distance and things based on it */
 845             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 846             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 847             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 848
 849             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 850             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 851             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 852
 853             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 854             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 855             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 856
 857             /* Load parameters for j particles */
 858             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 859
 860             fjx0             = _fjsp_setzero_v2r8();
 861             fjy0             = _fjsp_setzero_v2r8();
 862             fjz0             = _fjsp_setzero_v2r8();
 863
 864             /**************************
 865              * CALCULATE INTERACTIONS *
 866              **************************/
 867
 868             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 869             {
 870
 871             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 872
 873             /* Compute parameters for interactions between i and j atoms */
 874             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 875
 876             /* EWALD ELECTROSTATICS */
 877
 878             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 879             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 880             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 881             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 882             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 883
 884             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 885             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 886             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 887             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 888             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 889             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 890             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 891             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 892             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 893             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 894
 895             d                = _fjsp_sub_v2r8(r00,rswitch);
 896             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 897             d2               = _fjsp_mul_v2r8(d,d);
 898             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 899
 900             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 901
 902             /* Evaluate switch function */
 903             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 904             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
 905             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 906
 907             fscal            = felec;
 908
 909             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 910
 911             /* Update vectorial force */
 912             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 913             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 914             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 915
 916             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 917             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 918             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 919
 920             }
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 927             {
 928
 929             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 930
 931             /* Compute parameters for interactions between i and j atoms */
 932             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 933
 934             /* EWALD ELECTROSTATICS */
 935
 936             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 937             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 938             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 939             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 940             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 941
 942             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 943             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 944             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 945             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 946             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 947             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 948             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 949             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 950             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 951             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 952
 953             d                = _fjsp_sub_v2r8(r10,rswitch);
 954             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 955             d2               = _fjsp_mul_v2r8(d,d);
 956             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 957
 958             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 959
 960             /* Evaluate switch function */
 961             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 962             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
 963             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 964
 965             fscal            = felec;
 966
 967             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 968
 969             /* Update vectorial force */
 970             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 971             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 972             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 973
 974             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 975             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 976             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 977
 978             }
 979
 980             /**************************
 981              * CALCULATE INTERACTIONS *
 982              **************************/
 983
 984             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 985             {
 986
 987             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 988
 989             /* Compute parameters for interactions between i and j atoms */
 990             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 991
 992             /* EWALD ELECTROSTATICS */
 993
 994             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 995             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 996             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 997             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 998             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 999
1000             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1001             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
1002             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1003             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1004             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
1005             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1006             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1007             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1008             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1009             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1010
1011             d                = _fjsp_sub_v2r8(r20,rswitch);
1012             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1013             d2               = _fjsp_mul_v2r8(d,d);
1014             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1015
1016             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1017
1018             /* Evaluate switch function */
1019             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1020             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1021             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1022
1023             fscal            = felec;
1024
1025             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1026
1027             /* Update vectorial force */
1028             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1029             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1030             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1031
1032             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1033             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1034             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1035
1036             }
1037
1038             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1039
1040             /* Inner loop uses 198 flops */
1041         }
1042
1043         if(jidx<j_index_end)
1044         {
1045
1046             jnrA             = jjnr[jidx];
1047             j_coord_offsetA  = DIM*jnrA;
1048
1049             /* load j atom coordinates */
1050             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1051                                               &jx0,&jy0,&jz0);
1052
1053             /* Calculate displacement vector */
1054             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1055             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1056             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1057             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1058             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1059             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1060             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1061             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1062             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1063
1064             /* Calculate squared distance and things based on it */
1065             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1066             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1067             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1068
1069             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1070             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1071             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1072
1073             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1074             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1075             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1076
1077             /* Load parameters for j particles */
1078             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1079
1080             fjx0             = _fjsp_setzero_v2r8();
1081             fjy0             = _fjsp_setzero_v2r8();
1082             fjz0             = _fjsp_setzero_v2r8();
1083
1084             /**************************
1085              * CALCULATE INTERACTIONS *
1086              **************************/
1087
1088             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1089             {
1090
1091             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1092
1093             /* Compute parameters for interactions between i and j atoms */
1094             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1095
1096             /* EWALD ELECTROSTATICS */
1097
1098             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1099             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1100             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1101             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1102             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1103
1104             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1105             ewtabD           = _fjsp_setzero_v2r8();
1106             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1107             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1108             ewtabFn          = _fjsp_setzero_v2r8();
1109             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1110             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1111             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1112             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
1113             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1114
1115             d                = _fjsp_sub_v2r8(r00,rswitch);
1116             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1117             d2               = _fjsp_mul_v2r8(d,d);
1118             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1119
1120             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1121
1122             /* Evaluate switch function */
1123             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1124             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(velec,dsw)) );
1125             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1126
1127             fscal            = felec;
1128
1129             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1130
1131             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1132
1133             /* Update vectorial force */
1134             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1135             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1136             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1137
1138             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1139             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1140             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1141
1142             }
1143
1144             /**************************
1145              * CALCULATE INTERACTIONS *
1146              **************************/
1147
1148             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1149             {
1150
1151             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1152
1153             /* Compute parameters for interactions between i and j atoms */
1154             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1155
1156             /* EWALD ELECTROSTATICS */
1157
1158             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1159             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1160             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1161             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1162             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1163
1164             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1165             ewtabD           = _fjsp_setzero_v2r8();
1166             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1167             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1168             ewtabFn          = _fjsp_setzero_v2r8();
1169             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1170             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1171             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1172             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
1173             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1174
1175             d                = _fjsp_sub_v2r8(r10,rswitch);
1176             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1177             d2               = _fjsp_mul_v2r8(d,d);
1178             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1179
1180             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1181
1182             /* Evaluate switch function */
1183             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1184             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv10,_fjsp_mul_v2r8(velec,dsw)) );
1185             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1186
1187             fscal            = felec;
1188
1189             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1190
1191             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1192
1193             /* Update vectorial force */
1194             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1195             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1196             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1197
1198             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1199             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1200             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1201
1202             }
1203
1204             /**************************
1205              * CALCULATE INTERACTIONS *
1206              **************************/
1207
1208             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1209             {
1210
1211             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1212
1213             /* Compute parameters for interactions between i and j atoms */
1214             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1215
1216             /* EWALD ELECTROSTATICS */
1217
1218             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1219             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1220             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1221             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1222             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1223
1224             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
1225             ewtabD           = _fjsp_setzero_v2r8();
1226             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
1227             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
1228             ewtabFn          = _fjsp_setzero_v2r8();
1229             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
1230             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
1231             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
1232             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
1233             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1234
1235             d                = _fjsp_sub_v2r8(r20,rswitch);
1236             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
1237             d2               = _fjsp_mul_v2r8(d,d);
1238             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
1239
1240             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
1241
1242             /* Evaluate switch function */
1243             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
1244             felec            = _fjsp_msub_v2r8( felec,sw , _fjsp_mul_v2r8(rinv20,_fjsp_mul_v2r8(velec,dsw)) );
1245             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1246
1247             fscal            = felec;
1248
1249             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1250
1251             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1252
1253             /* Update vectorial force */
1254             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1255             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1256             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1257
1258             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1259             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1260             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1261
1262             }
1263
1264             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1265
1266             /* Inner loop uses 198 flops */
1267         }
1268
1269         /* End of innermost loop */
1270
1271         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1272                                               f+i_coord_offset,fshift+i_shift_offset);
1273
1274         /* Increment number of inner iterations */
1275         inneriter                  += j_index_end - j_index_start;
1276
1277         /* Outer loop uses 18 flops */
1278     }
1279
1280     /* Increment number of outer iterations */
1281     outeriter        += nri;
1282
1283     /* Update outer/inner flops */
1284
1285     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_W3_F,outeriter*18 + inneriter*198);
1286 }