src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSh_VdwLJEwSh_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 "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            LJEwald
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEwSh_VdwLJEwSh_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              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8           c6grid_00;
 104     _fjsp_v2r8           c6grid_10;
 105     _fjsp_v2r8           c6grid_20;
 106     _fjsp_v2r8           c6grid_30;
 107     real                 *vdwgridparam;
 108     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 109     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 110     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 111     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 112     real             *ewtab;
 113     _fjsp_v2r8       itab_tmp;
 114     _fjsp_v2r8       dummy_mask,cutoff_mask;
 115     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 116     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 117     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 118
 119     x                = xx[0];
 120     f                = ff[0];
 121
 122     nri              = nlist->nri;
 123     iinr             = nlist->iinr;
 124     jindex           = nlist->jindex;
 125     jjnr             = nlist->jjnr;
 126     shiftidx         = nlist->shift;
 127     gid              = nlist->gid;
 128     shiftvec         = fr->shift_vec[0];
 129     fshift           = fr->fshift[0];
 130     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 131     charge           = mdatoms->chargeA;
 132     nvdwtype         = fr->ntype;
 133     vdwparam         = fr->nbfp;
 134     vdwtype          = mdatoms->typeA;
 135     vdwgridparam     = fr->ljpme_c6grid;
 136     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 137     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 138     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 139
 140     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 141     ewtab            = fr->ic->tabq_coul_FDV0;
 142     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 143     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 144
 145     /* Setup water-specific parameters */
 146     inr              = nlist->iinr[0];
 147     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 148     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 149     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 150     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 151
 152     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 153     rcutoff_scalar   = fr->rcoulomb;
 154     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 155     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 156
 157     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 158     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 159
 160     /* Avoid stupid compiler warnings */
 161     jnrA = jnrB = 0;
 162     j_coord_offsetA = 0;
 163     j_coord_offsetB = 0;
 164
 165     outeriter        = 0;
 166     inneriter        = 0;
 167
 168     /* Start outer loop over neighborlists */
 169     for(iidx=0; iidx<nri; iidx++)
 170     {
 171         /* Load shift vector for this list */
 172         i_shift_offset   = DIM*shiftidx[iidx];
 173
 174         /* Load limits for loop over neighbors */
 175         j_index_start    = jindex[iidx];
 176         j_index_end      = jindex[iidx+1];
 177
 178         /* Get outer coordinate index */
 179         inr              = iinr[iidx];
 180         i_coord_offset   = DIM*inr;
 181
 182         /* Load i particle coords and add shift vector */
 183         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 184                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 185
 186         fix0             = _fjsp_setzero_v2r8();
 187         fiy0             = _fjsp_setzero_v2r8();
 188         fiz0             = _fjsp_setzero_v2r8();
 189         fix1             = _fjsp_setzero_v2r8();
 190         fiy1             = _fjsp_setzero_v2r8();
 191         fiz1             = _fjsp_setzero_v2r8();
 192         fix2             = _fjsp_setzero_v2r8();
 193         fiy2             = _fjsp_setzero_v2r8();
 194         fiz2             = _fjsp_setzero_v2r8();
 195         fix3             = _fjsp_setzero_v2r8();
 196         fiy3             = _fjsp_setzero_v2r8();
 197         fiz3             = _fjsp_setzero_v2r8();
 198
 199         /* Reset potential sums */
 200         velecsum         = _fjsp_setzero_v2r8();
 201         vvdwsum          = _fjsp_setzero_v2r8();
 202
 203         /* Start inner kernel loop */
 204         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 205         {
 206
 207             /* Get j neighbor index, and coordinate index */
 208             jnrA             = jjnr[jidx];
 209             jnrB             = jjnr[jidx+1];
 210             j_coord_offsetA  = DIM*jnrA;
 211             j_coord_offsetB  = DIM*jnrB;
 212
 213             /* load j atom coordinates */
 214             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 215                                               &jx0,&jy0,&jz0);
 216
 217             /* Calculate displacement vector */
 218             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 219             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 220             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 221             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 222             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 223             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 224             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 225             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 226             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 227             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 228             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 229             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 230
 231             /* Calculate squared distance and things based on it */
 232             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 233             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 234             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 235             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 236
 237             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 238             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 239             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 240             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 241
 242             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 243             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 244             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 245             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 246
 247             /* Load parameters for j particles */
 248             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 249             vdwjidx0A        = 2*vdwtype[jnrA+0];
 250             vdwjidx0B        = 2*vdwtype[jnrB+0];
 251
 252             fjx0             = _fjsp_setzero_v2r8();
 253             fjy0             = _fjsp_setzero_v2r8();
 254             fjz0             = _fjsp_setzero_v2r8();
 255
 256             /**************************
 257              * CALCULATE INTERACTIONS *
 258              **************************/
 259
 260             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 261             {
 262
 263             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 264
 265             /* Compute parameters for interactions between i and j atoms */
 266             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 267                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 268
 269             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 270                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 271
 272             /* Analytical LJ-PME */
 273             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 274             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 275             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 276             exponent         = gmx_simd_exp_d(-ewcljrsq);
 277             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 278             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 279             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 280             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
 281             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 282             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 283                                _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw6,_fjsp_madd_v2r8(c6grid_00,sh_lj_ewald,_fjsp_mul_v2r8(c6_00,sh_vdw_invrcut6))),one_sixth));
 284             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 285             fvdw             = _fjsp_mul_v2r8(_fjsp_add_v2r8(vvdw12,_fjsp_msub_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6),vvdw6)),rinvsq00);
 286
 287             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 288
 289             /* Update potential sum for this i atom from the interaction with this j atom. */
 290             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 291             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 292
 293             fscal            = fvdw;
 294
 295             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 296
 297             /* Update vectorial force */
 298             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 299             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 300             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 301
 302             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 303             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 304             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 305
 306             }
 307
 308             /**************************
 309              * CALCULATE INTERACTIONS *
 310              **************************/
 311
 312             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 313             {
 314
 315             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 316
 317             /* Compute parameters for interactions between i and j atoms */
 318             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 319
 320             /* EWALD ELECTROSTATICS */
 321
 322             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 323             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 324             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 325             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 326             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 327
 328             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 329             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 330             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 331             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 332             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 333             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 334             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 335             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 336             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 337             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 338
 339             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 340
 341             /* Update potential sum for this i atom from the interaction with this j atom. */
 342             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 343             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 344
 345             fscal            = felec;
 346
 347             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 348
 349             /* Update vectorial force */
 350             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 351             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 352             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 353
 354             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 355             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 356             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 357
 358             }
 359
 360             /**************************
 361              * CALCULATE INTERACTIONS *
 362              **************************/
 363
 364             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 365             {
 366
 367             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 368
 369             /* Compute parameters for interactions between i and j atoms */
 370             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 371
 372             /* EWALD ELECTROSTATICS */
 373
 374             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 375             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 376             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 377             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 378             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 379
 380             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 381             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 382             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 383             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 384             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 385             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 386             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 387             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 388             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 389             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 390
 391             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 392
 393             /* Update potential sum for this i atom from the interaction with this j atom. */
 394             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 395             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 396
 397             fscal            = felec;
 398
 399             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 400
 401             /* Update vectorial force */
 402             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 403             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 404             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 405
 406             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 407             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 408             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 409
 410             }
 411
 412             /**************************
 413              * CALCULATE INTERACTIONS *
 414              **************************/
 415
 416             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 417             {
 418
 419             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 420
 421             /* Compute parameters for interactions between i and j atoms */
 422             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 423
 424             /* EWALD ELECTROSTATICS */
 425
 426             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 427             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 428             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 429             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 430             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 431
 432             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 433             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 434             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 435             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 436             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 437             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 438             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 439             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 440             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
 441             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 442
 443             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 444
 445             /* Update potential sum for this i atom from the interaction with this j atom. */
 446             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 447             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 448
 449             fscal            = felec;
 450
 451             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 452
 453             /* Update vectorial force */
 454             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 455             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 456             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 457
 458             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 459             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 460             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 461
 462             }
 463
 464             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 465
 466             /* Inner loop uses 209 flops */
 467         }
 468
 469         if(jidx<j_index_end)
 470         {
 471
 472             jnrA             = jjnr[jidx];
 473             j_coord_offsetA  = DIM*jnrA;
 474
 475             /* load j atom coordinates */
 476             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 477                                               &jx0,&jy0,&jz0);
 478
 479             /* Calculate displacement vector */
 480             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 481             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 482             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 483             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 484             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 485             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 486             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 487             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 488             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 489             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 490             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 491             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 492
 493             /* Calculate squared distance and things based on it */
 494             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 495             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 496             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 497             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 498
 499             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 500             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 501             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 502             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 503
 504             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 505             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 506             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 507             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 508
 509             /* Load parameters for j particles */
 510             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 511             vdwjidx0A        = 2*vdwtype[jnrA+0];
 512
 513             fjx0             = _fjsp_setzero_v2r8();
 514             fjy0             = _fjsp_setzero_v2r8();
 515             fjz0             = _fjsp_setzero_v2r8();
 516
 517             /**************************
 518              * CALCULATE INTERACTIONS *
 519              **************************/
 520
 521             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 522             {
 523
 524             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 525
 526             /* Compute parameters for interactions between i and j atoms */
 527             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 528
 529             c6grid_00       = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
 530
 531             /* Analytical LJ-PME */
 532             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 533             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 534             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 535             exponent         = gmx_simd_exp_d(-ewcljrsq);
 536             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 537             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 538             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 539             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
 540             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 541             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 542                                _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw6,_fjsp_madd_v2r8(c6grid_00,sh_lj_ewald,_fjsp_mul_v2r8(c6_00,sh_vdw_invrcut6))),one_sixth));
 543             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 544             fvdw             = _fjsp_mul_v2r8(_fjsp_add_v2r8(vvdw12,_fjsp_msub_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6),vvdw6)),rinvsq00);
 545
 546             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 547
 548             /* Update potential sum for this i atom from the interaction with this j atom. */
 549             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 550             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 551             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 552
 553             fscal            = fvdw;
 554
 555             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 556
 557             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 558
 559             /* Update vectorial force */
 560             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 561             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 562             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 563
 564             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 565             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 566             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 567
 568             }
 569
 570             /**************************
 571              * CALCULATE INTERACTIONS *
 572              **************************/
 573
 574             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 575             {
 576
 577             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 578
 579             /* Compute parameters for interactions between i and j atoms */
 580             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 581
 582             /* EWALD ELECTROSTATICS */
 583
 584             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 585             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 586             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 587             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 588             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 589
 590             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 591             ewtabD           = _fjsp_setzero_v2r8();
 592             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 593             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 594             ewtabFn          = _fjsp_setzero_v2r8();
 595             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 596             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 597             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 598             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 599             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 600
 601             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 602
 603             /* Update potential sum for this i atom from the interaction with this j atom. */
 604             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 605             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 606             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 607
 608             fscal            = felec;
 609
 610             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 611
 612             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 613
 614             /* Update vectorial force */
 615             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 616             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 617             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 618
 619             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 620             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 621             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 622
 623             }
 624
 625             /**************************
 626              * CALCULATE INTERACTIONS *
 627              **************************/
 628
 629             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 630             {
 631
 632             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 633
 634             /* Compute parameters for interactions between i and j atoms */
 635             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 636
 637             /* EWALD ELECTROSTATICS */
 638
 639             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 640             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 641             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 642             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 643             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 644
 645             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 646             ewtabD           = _fjsp_setzero_v2r8();
 647             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 648             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 649             ewtabFn          = _fjsp_setzero_v2r8();
 650             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 651             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 652             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 653             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 654             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 655
 656             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 657
 658             /* Update potential sum for this i atom from the interaction with this j atom. */
 659             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 660             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 661             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 662
 663             fscal            = felec;
 664
 665             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 666
 667             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 668
 669             /* Update vectorial force */
 670             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 671             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 672             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 673
 674             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 675             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 676             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 677
 678             }
 679
 680             /**************************
 681              * CALCULATE INTERACTIONS *
 682              **************************/
 683
 684             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 685             {
 686
 687             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 688
 689             /* Compute parameters for interactions between i and j atoms */
 690             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 691
 692             /* EWALD ELECTROSTATICS */
 693
 694             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 695             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 696             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 697             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 698             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 699
 700             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 701             ewtabD           = _fjsp_setzero_v2r8();
 702             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 703             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 704             ewtabFn          = _fjsp_setzero_v2r8();
 705             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 706             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 707             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 708             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
 709             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 710
 711             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 712
 713             /* Update potential sum for this i atom from the interaction with this j atom. */
 714             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 715             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 716             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 717
 718             fscal            = felec;
 719
 720             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 721
 722             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 723
 724             /* Update vectorial force */
 725             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 726             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 727             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 728
 729             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 730             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 731             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 732
 733             }
 734
 735             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 736
 737             /* Inner loop uses 209 flops */
 738         }
 739
 740         /* End of innermost loop */
 741
 742         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 743                                               f+i_coord_offset,fshift+i_shift_offset);
 744
 745         ggid                        = gid[iidx];
 746         /* Update potential energies */
 747         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 748         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 749
 750         /* Increment number of inner iterations */
 751         inneriter                  += j_index_end - j_index_start;
 752
 753         /* Outer loop uses 26 flops */
 754     }
 755
 756     /* Increment number of outer iterations */
 757     outeriter        += nri;
 758
 759     /* Update outer/inner flops */
 760
 761     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*209);
 762 }
 763 /*
 764  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_sparc64_hpc_ace_double
 765  * Electrostatics interaction: Ewald
 766  * VdW interaction:            LJEwald
 767  * Geometry:                   Water4-Particle
 768  * Calculate force/pot:        Force
 769  */
 770 void
 771 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_sparc64_hpc_ace_double
 772                     (t_nblist                    * gmx_restrict       nlist,
 773                      rvec                        * gmx_restrict          xx,
 774                      rvec                        * gmx_restrict          ff,
 775                      t_forcerec                  * gmx_restrict          fr,
 776                      t_mdatoms                   * gmx_restrict     mdatoms,
 777                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 778                      t_nrnb                      * gmx_restrict        nrnb)
 779 {
 780     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 781      * just 0 for non-waters.
 782      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 783      * jnr indices corresponding to data put in the four positions in the SIMD register.
 784      */
 785     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 786     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 787     int              jnrA,jnrB;
 788     int              j_coord_offsetA,j_coord_offsetB;
 789     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 790     real             rcutoff_scalar;
 791     real             *shiftvec,*fshift,*x,*f;
 792     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 793     int              vdwioffset0;
 794     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 795     int              vdwioffset1;
 796     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 797     int              vdwioffset2;
 798     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 799     int              vdwioffset3;
 800     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 801     int              vdwjidx0A,vdwjidx0B;
 802     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 803     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 804     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 805     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 806     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 807     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 808     real             *charge;
 809     int              nvdwtype;
 810     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 811     int              *vdwtype;
 812     real             *vdwparam;
 813     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 814     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 815     _fjsp_v2r8           c6grid_00;
 816     _fjsp_v2r8           c6grid_10;
 817     _fjsp_v2r8           c6grid_20;
 818     _fjsp_v2r8           c6grid_30;
 819     real                 *vdwgridparam;
 820     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 821     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 822     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 823     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 824     real             *ewtab;
 825     _fjsp_v2r8       itab_tmp;
 826     _fjsp_v2r8       dummy_mask,cutoff_mask;
 827     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 828     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 829     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 830
 831     x                = xx[0];
 832     f                = ff[0];
 833
 834     nri              = nlist->nri;
 835     iinr             = nlist->iinr;
 836     jindex           = nlist->jindex;
 837     jjnr             = nlist->jjnr;
 838     shiftidx         = nlist->shift;
 839     gid              = nlist->gid;
 840     shiftvec         = fr->shift_vec[0];
 841     fshift           = fr->fshift[0];
 842     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 843     charge           = mdatoms->chargeA;
 844     nvdwtype         = fr->ntype;
 845     vdwparam         = fr->nbfp;
 846     vdwtype          = mdatoms->typeA;
 847     vdwgridparam     = fr->ljpme_c6grid;
 848     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 849     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 850     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 851
 852     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 853     ewtab            = fr->ic->tabq_coul_F;
 854     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 855     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 856
 857     /* Setup water-specific parameters */
 858     inr              = nlist->iinr[0];
 859     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 860     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 861     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 862     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 863
 864     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 865     rcutoff_scalar   = fr->rcoulomb;
 866     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 867     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 868
 869     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 870     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 871
 872     /* Avoid stupid compiler warnings */
 873     jnrA = jnrB = 0;
 874     j_coord_offsetA = 0;
 875     j_coord_offsetB = 0;
 876
 877     outeriter        = 0;
 878     inneriter        = 0;
 879
 880     /* Start outer loop over neighborlists */
 881     for(iidx=0; iidx<nri; iidx++)
 882     {
 883         /* Load shift vector for this list */
 884         i_shift_offset   = DIM*shiftidx[iidx];
 885
 886         /* Load limits for loop over neighbors */
 887         j_index_start    = jindex[iidx];
 888         j_index_end      = jindex[iidx+1];
 889
 890         /* Get outer coordinate index */
 891         inr              = iinr[iidx];
 892         i_coord_offset   = DIM*inr;
 893
 894         /* Load i particle coords and add shift vector */
 895         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 896                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 897
 898         fix0             = _fjsp_setzero_v2r8();
 899         fiy0             = _fjsp_setzero_v2r8();
 900         fiz0             = _fjsp_setzero_v2r8();
 901         fix1             = _fjsp_setzero_v2r8();
 902         fiy1             = _fjsp_setzero_v2r8();
 903         fiz1             = _fjsp_setzero_v2r8();
 904         fix2             = _fjsp_setzero_v2r8();
 905         fiy2             = _fjsp_setzero_v2r8();
 906         fiz2             = _fjsp_setzero_v2r8();
 907         fix3             = _fjsp_setzero_v2r8();
 908         fiy3             = _fjsp_setzero_v2r8();
 909         fiz3             = _fjsp_setzero_v2r8();
 910
 911         /* Start inner kernel loop */
 912         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 913         {
 914
 915             /* Get j neighbor index, and coordinate index */
 916             jnrA             = jjnr[jidx];
 917             jnrB             = jjnr[jidx+1];
 918             j_coord_offsetA  = DIM*jnrA;
 919             j_coord_offsetB  = DIM*jnrB;
 920
 921             /* load j atom coordinates */
 922             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 923                                               &jx0,&jy0,&jz0);
 924
 925             /* Calculate displacement vector */
 926             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 927             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 928             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 929             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 930             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 931             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 932             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 933             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 934             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 935             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 936             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 937             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 938
 939             /* Calculate squared distance and things based on it */
 940             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 941             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 942             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 943             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 944
 945             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 946             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 947             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 948             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 949
 950             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 951             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 952             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 953             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 954
 955             /* Load parameters for j particles */
 956             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 957             vdwjidx0A        = 2*vdwtype[jnrA+0];
 958             vdwjidx0B        = 2*vdwtype[jnrB+0];
 959
 960             fjx0             = _fjsp_setzero_v2r8();
 961             fjy0             = _fjsp_setzero_v2r8();
 962             fjz0             = _fjsp_setzero_v2r8();
 963
 964             /**************************
 965              * CALCULATE INTERACTIONS *
 966              **************************/
 967
 968             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 969             {
 970
 971             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 972
 973             /* Compute parameters for interactions between i and j atoms */
 974             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 975                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 976
 977             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 978                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 979
 980             /* Analytical LJ-PME */
 981             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 982             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 983             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 984             exponent         = gmx_simd_exp_d(-ewcljrsq);
 985             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 986             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 987             /* f6A = 6 * C6grid * (1 - poly) */
 988             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
 989             /* f6B = C6grid * exponent * beta^6 */
 990             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
 991             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 992             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
 993
 994             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 995
 996             fscal            = fvdw;
 997
 998             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 999
1000             /* Update vectorial force */
1001             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1002             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1003             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1004
1005             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1006             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1007             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1008
1009             }
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1016             {
1017
1018             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1019
1020             /* Compute parameters for interactions between i and j atoms */
1021             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1022
1023             /* EWALD ELECTROSTATICS */
1024
1025             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1026             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1027             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1028             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1029             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1030
1031             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1032                                          &ewtabF,&ewtabFn);
1033             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1034             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1035
1036             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1037
1038             fscal            = felec;
1039
1040             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1041
1042             /* Update vectorial force */
1043             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1044             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1045             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1046
1047             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1048             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1049             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1050
1051             }
1052
1053             /**************************
1054              * CALCULATE INTERACTIONS *
1055              **************************/
1056
1057             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1058             {
1059
1060             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1061
1062             /* Compute parameters for interactions between i and j atoms */
1063             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1064
1065             /* EWALD ELECTROSTATICS */
1066
1067             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1068             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1069             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1070             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1071             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1072
1073             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1074                                          &ewtabF,&ewtabFn);
1075             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1076             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1077
1078             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1079
1080             fscal            = felec;
1081
1082             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1083
1084             /* Update vectorial force */
1085             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1086             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1087             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1088
1089             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1090             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1091             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1092
1093             }
1094
1095             /**************************
1096              * CALCULATE INTERACTIONS *
1097              **************************/
1098
1099             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1100             {
1101
1102             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1103
1104             /* Compute parameters for interactions between i and j atoms */
1105             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1106
1107             /* EWALD ELECTROSTATICS */
1108
1109             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1110             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1111             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1112             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1113             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1114
1115             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1116                                          &ewtabF,&ewtabFn);
1117             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1118             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1119
1120             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1121
1122             fscal            = felec;
1123
1124             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1125
1126             /* Update vectorial force */
1127             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1128             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1129             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1130
1131             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1132             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1133             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1134
1135             }
1136
1137             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1138
1139             /* Inner loop uses 180 flops */
1140         }
1141
1142         if(jidx<j_index_end)
1143         {
1144
1145             jnrA             = jjnr[jidx];
1146             j_coord_offsetA  = DIM*jnrA;
1147
1148             /* load j atom coordinates */
1149             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1150                                               &jx0,&jy0,&jz0);
1151
1152             /* Calculate displacement vector */
1153             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1154             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1155             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1156             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1157             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1158             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1159             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1160             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1161             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1162             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1163             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1164             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1165
1166             /* Calculate squared distance and things based on it */
1167             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1168             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1169             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1170             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1171
1172             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1173             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1174             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1175             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1176
1177             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1178             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1179             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1180             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1181
1182             /* Load parameters for j particles */
1183             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1184             vdwjidx0A        = 2*vdwtype[jnrA+0];
1185
1186             fjx0             = _fjsp_setzero_v2r8();
1187             fjy0             = _fjsp_setzero_v2r8();
1188             fjz0             = _fjsp_setzero_v2r8();
1189
1190             /**************************
1191              * CALCULATE INTERACTIONS *
1192              **************************/
1193
1194             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1195             {
1196
1197             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1198
1199             /* Compute parameters for interactions between i and j atoms */
1200             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1201
1202             c6grid_00       = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
1203
1204             /* Analytical LJ-PME */
1205             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1206             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
1207             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
1208             exponent         = gmx_simd_exp_d(-ewcljrsq);
1209             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1210             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
1211             /* f6A = 6 * C6grid * (1 - poly) */
1212             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
1213             /* f6B = C6grid * exponent * beta^6 */
1214             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
1215             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1216             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1217
1218             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1219
1220             fscal            = fvdw;
1221
1222             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1223
1224             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1225
1226             /* Update vectorial force */
1227             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1228             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1229             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1230
1231             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1232             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1233             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1234
1235             }
1236
1237             /**************************
1238              * CALCULATE INTERACTIONS *
1239              **************************/
1240
1241             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1242             {
1243
1244             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1245
1246             /* Compute parameters for interactions between i and j atoms */
1247             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1248
1249             /* EWALD ELECTROSTATICS */
1250
1251             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1252             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1253             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1254             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1255             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1256
1257             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1258             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1259             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1260
1261             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1262
1263             fscal            = felec;
1264
1265             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1266
1267             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1268
1269             /* Update vectorial force */
1270             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1271             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1272             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1273
1274             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1275             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1276             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1277
1278             }
1279
1280             /**************************
1281              * CALCULATE INTERACTIONS *
1282              **************************/
1283
1284             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1285             {
1286
1287             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1288
1289             /* Compute parameters for interactions between i and j atoms */
1290             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1291
1292             /* EWALD ELECTROSTATICS */
1293
1294             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1295             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1296             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1297             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1298             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1299
1300             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1301             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1302             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1303
1304             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1305
1306             fscal            = felec;
1307
1308             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1309
1310             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1311
1312             /* Update vectorial force */
1313             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1314             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1315             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1316
1317             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1318             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1319             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1320
1321             }
1322
1323             /**************************
1324              * CALCULATE INTERACTIONS *
1325              **************************/
1326
1327             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1328             {
1329
1330             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1331
1332             /* Compute parameters for interactions between i and j atoms */
1333             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1334
1335             /* EWALD ELECTROSTATICS */
1336
1337             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1338             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1339             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1340             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1341             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1342
1343             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1344             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1345             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1346
1347             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1348
1349             fscal            = felec;
1350
1351             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1352
1353             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1354
1355             /* Update vectorial force */
1356             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1357             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1358             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1359
1360             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1361             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1362             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1363
1364             }
1365
1366             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1367
1368             /* Inner loop uses 180 flops */
1369         }
1370
1371         /* End of innermost loop */
1372
1373         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1374                                               f+i_coord_offset,fshift+i_shift_offset);
1375
1376         /* Increment number of inner iterations */
1377         inneriter                  += j_index_end - j_index_start;
1378
1379         /* Outer loop uses 24 flops */
1380     }
1381
1382     /* Increment number of outer iterations */
1383     outeriter        += nri;
1384
1385     /* Update outer/inner flops */
1386
1387     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*180);
1388 }