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 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/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(poly,one),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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 528                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 529
 530             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 531                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 532
 533             /* Analytical LJ-PME */
 534             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 535             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 536             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 537             exponent         = gmx_simd_exp_d(ewcljrsq);
 538             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 539             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 540             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 541             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(c6grid_00,_fjsp_sub_v2r8(poly,one),c6_00),rinvsix);
 542             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 543             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 544                                _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));
 545             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 546             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);
 547
 548             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 549
 550             /* Update potential sum for this i atom from the interaction with this j atom. */
 551             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 552             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 553             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 554
 555             fscal            = fvdw;
 556
 557             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 558
 559             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 560
 561             /* Update vectorial force */
 562             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 563             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 564             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 565
 566             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 567             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 568             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 569
 570             }
 571
 572             /**************************
 573              * CALCULATE INTERACTIONS *
 574              **************************/
 575
 576             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 577             {
 578
 579             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 580
 581             /* Compute parameters for interactions between i and j atoms */
 582             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 583
 584             /* EWALD ELECTROSTATICS */
 585
 586             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 587             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 588             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 589             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 590             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 591
 592             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 593             ewtabD           = _fjsp_setzero_v2r8();
 594             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 595             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 596             ewtabFn          = _fjsp_setzero_v2r8();
 597             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 598             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 599             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 600             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 601             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 602
 603             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 604
 605             /* Update potential sum for this i atom from the interaction with this j atom. */
 606             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 607             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 608             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 609
 610             fscal            = felec;
 611
 612             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 613
 614             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 615
 616             /* Update vectorial force */
 617             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 618             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 619             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 620
 621             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 622             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 623             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 624
 625             }
 626
 627             /**************************
 628              * CALCULATE INTERACTIONS *
 629              **************************/
 630
 631             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 632             {
 633
 634             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 635
 636             /* Compute parameters for interactions between i and j atoms */
 637             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 638
 639             /* EWALD ELECTROSTATICS */
 640
 641             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 642             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 643             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 644             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 645             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 646
 647             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 648             ewtabD           = _fjsp_setzero_v2r8();
 649             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 650             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 651             ewtabFn          = _fjsp_setzero_v2r8();
 652             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 653             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 654             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 655             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 656             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 657
 658             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 659
 660             /* Update potential sum for this i atom from the interaction with this j atom. */
 661             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 662             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 663             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 664
 665             fscal            = felec;
 666
 667             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 668
 669             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 670
 671             /* Update vectorial force */
 672             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 673             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 674             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 675
 676             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 677             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 678             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 679
 680             }
 681
 682             /**************************
 683              * CALCULATE INTERACTIONS *
 684              **************************/
 685
 686             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 687             {
 688
 689             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 690
 691             /* Compute parameters for interactions between i and j atoms */
 692             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 693
 694             /* EWALD ELECTROSTATICS */
 695
 696             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 697             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 698             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 699             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 700             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 701
 702             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 703             ewtabD           = _fjsp_setzero_v2r8();
 704             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 705             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 706             ewtabFn          = _fjsp_setzero_v2r8();
 707             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 708             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 709             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 710             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
 711             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 712
 713             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 714
 715             /* Update potential sum for this i atom from the interaction with this j atom. */
 716             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 717             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 718             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 719
 720             fscal            = felec;
 721
 722             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 723
 724             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 725
 726             /* Update vectorial force */
 727             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 728             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 729             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 730
 731             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 732             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 733             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 734
 735             }
 736
 737             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 738
 739             /* Inner loop uses 209 flops */
 740         }
 741
 742         /* End of innermost loop */
 743
 744         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 745                                               f+i_coord_offset,fshift+i_shift_offset);
 746
 747         ggid                        = gid[iidx];
 748         /* Update potential energies */
 749         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 750         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 751
 752         /* Increment number of inner iterations */
 753         inneriter                  += j_index_end - j_index_start;
 754
 755         /* Outer loop uses 26 flops */
 756     }
 757
 758     /* Increment number of outer iterations */
 759     outeriter        += nri;
 760
 761     /* Update outer/inner flops */
 762
 763     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*209);
 764 }
 765 /*
 766  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_sparc64_hpc_ace_double
 767  * Electrostatics interaction: Ewald
 768  * VdW interaction:            LJEwald
 769  * Geometry:                   Water4-Particle
 770  * Calculate force/pot:        Force
 771  */
 772 void
 773 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_sparc64_hpc_ace_double
 774                     (t_nblist                    * gmx_restrict       nlist,
 775                      rvec                        * gmx_restrict          xx,
 776                      rvec                        * gmx_restrict          ff,
 777                      t_forcerec                  * gmx_restrict          fr,
 778                      t_mdatoms                   * gmx_restrict     mdatoms,
 779                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 780                      t_nrnb                      * gmx_restrict        nrnb)
 781 {
 782     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 783      * just 0 for non-waters.
 784      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 785      * jnr indices corresponding to data put in the four positions in the SIMD register.
 786      */
 787     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 788     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 789     int              jnrA,jnrB;
 790     int              j_coord_offsetA,j_coord_offsetB;
 791     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 792     real             rcutoff_scalar;
 793     real             *shiftvec,*fshift,*x,*f;
 794     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 795     int              vdwioffset0;
 796     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 797     int              vdwioffset1;
 798     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 799     int              vdwioffset2;
 800     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 801     int              vdwioffset3;
 802     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 803     int              vdwjidx0A,vdwjidx0B;
 804     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 805     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 806     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 807     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 808     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 809     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 810     real             *charge;
 811     int              nvdwtype;
 812     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 813     int              *vdwtype;
 814     real             *vdwparam;
 815     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 816     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 817     _fjsp_v2r8           c6grid_00;
 818     _fjsp_v2r8           c6grid_10;
 819     _fjsp_v2r8           c6grid_20;
 820     _fjsp_v2r8           c6grid_30;
 821     real                 *vdwgridparam;
 822     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 823     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 824     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 825     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 826     real             *ewtab;
 827     _fjsp_v2r8       itab_tmp;
 828     _fjsp_v2r8       dummy_mask,cutoff_mask;
 829     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 830     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 831     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 832
 833     x                = xx[0];
 834     f                = ff[0];
 835
 836     nri              = nlist->nri;
 837     iinr             = nlist->iinr;
 838     jindex           = nlist->jindex;
 839     jjnr             = nlist->jjnr;
 840     shiftidx         = nlist->shift;
 841     gid              = nlist->gid;
 842     shiftvec         = fr->shift_vec[0];
 843     fshift           = fr->fshift[0];
 844     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 845     charge           = mdatoms->chargeA;
 846     nvdwtype         = fr->ntype;
 847     vdwparam         = fr->nbfp;
 848     vdwtype          = mdatoms->typeA;
 849     vdwgridparam     = fr->ljpme_c6grid;
 850     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 851     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 852     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 853
 854     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 855     ewtab            = fr->ic->tabq_coul_F;
 856     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 857     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 858
 859     /* Setup water-specific parameters */
 860     inr              = nlist->iinr[0];
 861     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 862     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 863     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 864     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 865
 866     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 867     rcutoff_scalar   = fr->rcoulomb;
 868     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 869     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 870
 871     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 872     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 873
 874     /* Avoid stupid compiler warnings */
 875     jnrA = jnrB = 0;
 876     j_coord_offsetA = 0;
 877     j_coord_offsetB = 0;
 878
 879     outeriter        = 0;
 880     inneriter        = 0;
 881
 882     /* Start outer loop over neighborlists */
 883     for(iidx=0; iidx<nri; iidx++)
 884     {
 885         /* Load shift vector for this list */
 886         i_shift_offset   = DIM*shiftidx[iidx];
 887
 888         /* Load limits for loop over neighbors */
 889         j_index_start    = jindex[iidx];
 890         j_index_end      = jindex[iidx+1];
 891
 892         /* Get outer coordinate index */
 893         inr              = iinr[iidx];
 894         i_coord_offset   = DIM*inr;
 895
 896         /* Load i particle coords and add shift vector */
 897         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 898                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 899
 900         fix0             = _fjsp_setzero_v2r8();
 901         fiy0             = _fjsp_setzero_v2r8();
 902         fiz0             = _fjsp_setzero_v2r8();
 903         fix1             = _fjsp_setzero_v2r8();
 904         fiy1             = _fjsp_setzero_v2r8();
 905         fiz1             = _fjsp_setzero_v2r8();
 906         fix2             = _fjsp_setzero_v2r8();
 907         fiy2             = _fjsp_setzero_v2r8();
 908         fiz2             = _fjsp_setzero_v2r8();
 909         fix3             = _fjsp_setzero_v2r8();
 910         fiy3             = _fjsp_setzero_v2r8();
 911         fiz3             = _fjsp_setzero_v2r8();
 912
 913         /* Start inner kernel loop */
 914         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 915         {
 916
 917             /* Get j neighbor index, and coordinate index */
 918             jnrA             = jjnr[jidx];
 919             jnrB             = jjnr[jidx+1];
 920             j_coord_offsetA  = DIM*jnrA;
 921             j_coord_offsetB  = DIM*jnrB;
 922
 923             /* load j atom coordinates */
 924             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 925                                               &jx0,&jy0,&jz0);
 926
 927             /* Calculate displacement vector */
 928             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 929             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 930             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 931             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 932             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 933             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 934             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 935             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 936             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 937             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 938             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 939             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 940
 941             /* Calculate squared distance and things based on it */
 942             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 943             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 944             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 945             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 946
 947             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 948             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 949             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 950             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 951
 952             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 953             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 954             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 955             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 956
 957             /* Load parameters for j particles */
 958             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 959             vdwjidx0A        = 2*vdwtype[jnrA+0];
 960             vdwjidx0B        = 2*vdwtype[jnrB+0];
 961
 962             fjx0             = _fjsp_setzero_v2r8();
 963             fjy0             = _fjsp_setzero_v2r8();
 964             fjz0             = _fjsp_setzero_v2r8();
 965
 966             /**************************
 967              * CALCULATE INTERACTIONS *
 968              **************************/
 969
 970             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 971             {
 972
 973             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 974
 975             /* Compute parameters for interactions between i and j atoms */
 976             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 977                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 978
 979             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 980                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 981
 982             /* Analytical LJ-PME */
 983             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 984             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 985             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 986             exponent         = gmx_simd_exp_d(ewcljrsq);
 987             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 988             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 989             /* f6A = 6 * C6grid * (1 - poly) */
 990             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_sub_v2r8(one,poly));
 991             /* f6B = C6grid * exponent * beta^6 */
 992             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
 993             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 994             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
 995
 996             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 997
 998             fscal            = fvdw;
 999
1000             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1001
1002             /* Update vectorial force */
1003             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1004             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1005             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1006
1007             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1008             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1009             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1010
1011             }
1012
1013             /**************************
1014              * CALCULATE INTERACTIONS *
1015              **************************/
1016
1017             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1018             {
1019
1020             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1021
1022             /* Compute parameters for interactions between i and j atoms */
1023             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1024
1025             /* EWALD ELECTROSTATICS */
1026
1027             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1028             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1029             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1030             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1031             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1032
1033             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1034                                          &ewtabF,&ewtabFn);
1035             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1036             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1037
1038             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1039
1040             fscal            = felec;
1041
1042             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1043
1044             /* Update vectorial force */
1045             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1046             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1047             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1048
1049             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1050             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1051             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1052
1053             }
1054
1055             /**************************
1056              * CALCULATE INTERACTIONS *
1057              **************************/
1058
1059             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1060             {
1061
1062             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1063
1064             /* Compute parameters for interactions between i and j atoms */
1065             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1066
1067             /* EWALD ELECTROSTATICS */
1068
1069             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1070             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1071             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1072             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1073             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1074
1075             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1076                                          &ewtabF,&ewtabFn);
1077             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1078             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1079
1080             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1081
1082             fscal            = felec;
1083
1084             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1085
1086             /* Update vectorial force */
1087             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1088             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1089             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1090
1091             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1092             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1093             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1094
1095             }
1096
1097             /**************************
1098              * CALCULATE INTERACTIONS *
1099              **************************/
1100
1101             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1102             {
1103
1104             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1105
1106             /* Compute parameters for interactions between i and j atoms */
1107             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1108
1109             /* EWALD ELECTROSTATICS */
1110
1111             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1112             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1113             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1114             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1115             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1116
1117             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1118                                          &ewtabF,&ewtabFn);
1119             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1120             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1121
1122             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1123
1124             fscal            = felec;
1125
1126             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1127
1128             /* Update vectorial force */
1129             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1130             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1131             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1132
1133             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1134             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1135             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1136
1137             }
1138
1139             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1140
1141             /* Inner loop uses 180 flops */
1142         }
1143
1144         if(jidx<j_index_end)
1145         {
1146
1147             jnrA             = jjnr[jidx];
1148             j_coord_offsetA  = DIM*jnrA;
1149
1150             /* load j atom coordinates */
1151             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1152                                               &jx0,&jy0,&jz0);
1153
1154             /* Calculate displacement vector */
1155             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1156             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1157             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1158             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1159             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1160             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1161             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1162             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1163             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1164             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1165             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1166             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1167
1168             /* Calculate squared distance and things based on it */
1169             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1170             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1171             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1172             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1173
1174             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1175             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1176             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1177             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1178
1179             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1180             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1181             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1182             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1183
1184             /* Load parameters for j particles */
1185             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1186             vdwjidx0A        = 2*vdwtype[jnrA+0];
1187
1188             fjx0             = _fjsp_setzero_v2r8();
1189             fjy0             = _fjsp_setzero_v2r8();
1190             fjz0             = _fjsp_setzero_v2r8();
1191
1192             /**************************
1193              * CALCULATE INTERACTIONS *
1194              **************************/
1195
1196             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1197             {
1198
1199             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1200
1201             /* Compute parameters for interactions between i and j atoms */
1202             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1203                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1204
1205             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
1206                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
1207
1208             /* Analytical LJ-PME */
1209             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1210             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
1211             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
1212             exponent         = gmx_simd_exp_d(ewcljrsq);
1213             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1214             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
1215             /* f6A = 6 * C6grid * (1 - poly) */
1216             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_sub_v2r8(one,poly));
1217             /* f6B = C6grid * exponent * beta^6 */
1218             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
1219             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1220             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1221
1222             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1223
1224             fscal            = fvdw;
1225
1226             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1227
1228             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1229
1230             /* Update vectorial force */
1231             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1232             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1233             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1234
1235             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1236             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1237             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1238
1239             }
1240
1241             /**************************
1242              * CALCULATE INTERACTIONS *
1243              **************************/
1244
1245             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1246             {
1247
1248             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1249
1250             /* Compute parameters for interactions between i and j atoms */
1251             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1252
1253             /* EWALD ELECTROSTATICS */
1254
1255             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1256             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1257             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1258             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1259             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1260
1261             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1262             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1263             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1264
1265             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1266
1267             fscal            = felec;
1268
1269             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1270
1271             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1272
1273             /* Update vectorial force */
1274             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1275             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1276             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1277
1278             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1279             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1280             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1281
1282             }
1283
1284             /**************************
1285              * CALCULATE INTERACTIONS *
1286              **************************/
1287
1288             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1289             {
1290
1291             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1292
1293             /* Compute parameters for interactions between i and j atoms */
1294             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1295
1296             /* EWALD ELECTROSTATICS */
1297
1298             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1299             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1300             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1301             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1302             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1303
1304             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1305             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1306             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1307
1308             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1309
1310             fscal            = felec;
1311
1312             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1313
1314             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1315
1316             /* Update vectorial force */
1317             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1318             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1319             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1320
1321             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1322             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1323             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1324
1325             }
1326
1327             /**************************
1328              * CALCULATE INTERACTIONS *
1329              **************************/
1330
1331             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1332             {
1333
1334             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1335
1336             /* Compute parameters for interactions between i and j atoms */
1337             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1338
1339             /* EWALD ELECTROSTATICS */
1340
1341             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1342             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1343             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1344             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1345             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1346
1347             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1348             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1349             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1350
1351             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1352
1353             fscal            = felec;
1354
1355             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1356
1357             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1358
1359             /* Update vectorial force */
1360             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1361             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1362             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1363
1364             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1365             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1366             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1367
1368             }
1369
1370             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1371
1372             /* Inner loop uses 180 flops */
1373         }
1374
1375         /* End of innermost loop */
1376
1377         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1378                                               f+i_coord_offset,fshift+i_shift_offset);
1379
1380         /* Increment number of inner iterations */
1381         inneriter                  += j_index_end - j_index_start;
1382
1383         /* Outer loop uses 24 flops */
1384     }
1385
1386     /* Increment number of outer iterations */
1387     outeriter        += nri;
1388
1389     /* Update outer/inner flops */
1390
1391     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*180);
1392 }