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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LJEwald
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8           c6grid_00;
 102     _fjsp_v2r8           c6grid_10;
 103     _fjsp_v2r8           c6grid_20;
 104     _fjsp_v2r8           c6grid_30;
 105     real                 *vdwgridparam;
 106     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 107     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 108     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 109     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 110     real             *ewtab;
 111     _fjsp_v2r8       itab_tmp;
 112     _fjsp_v2r8       dummy_mask,cutoff_mask;
 113     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 114     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 115     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 116
 117     x                = xx[0];
 118     f                = ff[0];
 119
 120     nri              = nlist->nri;
 121     iinr             = nlist->iinr;
 122     jindex           = nlist->jindex;
 123     jjnr             = nlist->jjnr;
 124     shiftidx         = nlist->shift;
 125     gid              = nlist->gid;
 126     shiftvec         = fr->shift_vec[0];
 127     fshift           = fr->fshift[0];
 128     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 129     charge           = mdatoms->chargeA;
 130     nvdwtype         = fr->ntype;
 131     vdwparam         = fr->nbfp;
 132     vdwtype          = mdatoms->typeA;
 133     vdwgridparam     = fr->ljpme_c6grid;
 134     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 135     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 136     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 137
 138     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 139     ewtab            = fr->ic->tabq_coul_FDV0;
 140     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 141     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 142
 143     /* Setup water-specific parameters */
 144     inr              = nlist->iinr[0];
 145     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 146     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 147     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 148     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 149
 150     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 151     rcutoff_scalar   = fr->rcoulomb;
 152     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 153     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 154
 155     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 156     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 157
 158     /* Avoid stupid compiler warnings */
 159     jnrA = jnrB = 0;
 160     j_coord_offsetA = 0;
 161     j_coord_offsetB = 0;
 162
 163     outeriter        = 0;
 164     inneriter        = 0;
 165
 166     /* Start outer loop over neighborlists */
 167     for(iidx=0; iidx<nri; iidx++)
 168     {
 169         /* Load shift vector for this list */
 170         i_shift_offset   = DIM*shiftidx[iidx];
 171
 172         /* Load limits for loop over neighbors */
 173         j_index_start    = jindex[iidx];
 174         j_index_end      = jindex[iidx+1];
 175
 176         /* Get outer coordinate index */
 177         inr              = iinr[iidx];
 178         i_coord_offset   = DIM*inr;
 179
 180         /* Load i particle coords and add shift vector */
 181         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 182                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 183
 184         fix0             = _fjsp_setzero_v2r8();
 185         fiy0             = _fjsp_setzero_v2r8();
 186         fiz0             = _fjsp_setzero_v2r8();
 187         fix1             = _fjsp_setzero_v2r8();
 188         fiy1             = _fjsp_setzero_v2r8();
 189         fiz1             = _fjsp_setzero_v2r8();
 190         fix2             = _fjsp_setzero_v2r8();
 191         fiy2             = _fjsp_setzero_v2r8();
 192         fiz2             = _fjsp_setzero_v2r8();
 193         fix3             = _fjsp_setzero_v2r8();
 194         fiy3             = _fjsp_setzero_v2r8();
 195         fiz3             = _fjsp_setzero_v2r8();
 196
 197         /* Reset potential sums */
 198         velecsum         = _fjsp_setzero_v2r8();
 199         vvdwsum          = _fjsp_setzero_v2r8();
 200
 201         /* Start inner kernel loop */
 202         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 203         {
 204
 205             /* Get j neighbor index, and coordinate index */
 206             jnrA             = jjnr[jidx];
 207             jnrB             = jjnr[jidx+1];
 208             j_coord_offsetA  = DIM*jnrA;
 209             j_coord_offsetB  = DIM*jnrB;
 210
 211             /* load j atom coordinates */
 212             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 213                                               &jx0,&jy0,&jz0);
 214
 215             /* Calculate displacement vector */
 216             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 217             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 218             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 219             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 220             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 221             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 222             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 223             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 224             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 225             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 226             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 227             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 228
 229             /* Calculate squared distance and things based on it */
 230             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 231             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 232             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 233             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 234
 235             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 236             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 237             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 238             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 239
 240             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 241             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 242             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 243             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 244
 245             /* Load parameters for j particles */
 246             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 247             vdwjidx0A        = 2*vdwtype[jnrA+0];
 248             vdwjidx0B        = 2*vdwtype[jnrB+0];
 249
 250             fjx0             = _fjsp_setzero_v2r8();
 251             fjy0             = _fjsp_setzero_v2r8();
 252             fjz0             = _fjsp_setzero_v2r8();
 253
 254             /**************************
 255              * CALCULATE INTERACTIONS *
 256              **************************/
 257
 258             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 259             {
 260
 261             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 262
 263             /* Compute parameters for interactions between i and j atoms */
 264             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 265                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 266
 267             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 268                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 269
 270             /* Analytical LJ-PME */
 271             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 272             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 273             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 274             exponent         = gmx_simd_exp_d(-ewcljrsq);
 275             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 276             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 277             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 278             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
 279             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 280             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 281                                _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));
 282             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 283             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);
 284
 285             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 286
 287             /* Update potential sum for this i atom from the interaction with this j atom. */
 288             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 289             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 290
 291             fscal            = fvdw;
 292
 293             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 294
 295             /* Update vectorial force */
 296             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 297             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 298             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 299
 300             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 301             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 302             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 303
 304             }
 305
 306             /**************************
 307              * CALCULATE INTERACTIONS *
 308              **************************/
 309
 310             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 311             {
 312
 313             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 314
 315             /* Compute parameters for interactions between i and j atoms */
 316             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 317
 318             /* EWALD ELECTROSTATICS */
 319
 320             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 321             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 322             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 323             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 324             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 325
 326             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 327             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 328             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 329             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 330             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 331             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 332             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 333             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 334             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 335             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 336
 337             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 338
 339             /* Update potential sum for this i atom from the interaction with this j atom. */
 340             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 341             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 342
 343             fscal            = felec;
 344
 345             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 346
 347             /* Update vectorial force */
 348             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 349             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 350             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 351
 352             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 353             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 354             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 355
 356             }
 357
 358             /**************************
 359              * CALCULATE INTERACTIONS *
 360              **************************/
 361
 362             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 363             {
 364
 365             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 366
 367             /* Compute parameters for interactions between i and j atoms */
 368             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 369
 370             /* EWALD ELECTROSTATICS */
 371
 372             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 373             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 374             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 375             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 376             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 377
 378             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 379             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 380             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 381             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 382             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 383             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 384             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 385             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 386             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 387             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 388
 389             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 390
 391             /* Update potential sum for this i atom from the interaction with this j atom. */
 392             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 393             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 394
 395             fscal            = felec;
 396
 397             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 398
 399             /* Update vectorial force */
 400             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 401             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 402             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 403
 404             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 405             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 406             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 407
 408             }
 409
 410             /**************************
 411              * CALCULATE INTERACTIONS *
 412              **************************/
 413
 414             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 415             {
 416
 417             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 418
 419             /* Compute parameters for interactions between i and j atoms */
 420             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 421
 422             /* EWALD ELECTROSTATICS */
 423
 424             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 425             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 426             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 427             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 428             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 429
 430             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 431             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 432             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 433             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 434             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 435             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 436             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 437             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 438             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
 439             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 440
 441             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 442
 443             /* Update potential sum for this i atom from the interaction with this j atom. */
 444             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 445             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 446
 447             fscal            = felec;
 448
 449             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 450
 451             /* Update vectorial force */
 452             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 453             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 454             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 455
 456             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 457             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 458             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 459
 460             }
 461
 462             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 463
 464             /* Inner loop uses 209 flops */
 465         }
 466
 467         if(jidx<j_index_end)
 468         {
 469
 470             jnrA             = jjnr[jidx];
 471             j_coord_offsetA  = DIM*jnrA;
 472
 473             /* load j atom coordinates */
 474             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 475                                               &jx0,&jy0,&jz0);
 476
 477             /* Calculate displacement vector */
 478             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 479             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 480             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 481             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 482             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 483             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 484             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 485             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 486             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 487             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 488             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 489             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 490
 491             /* Calculate squared distance and things based on it */
 492             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 493             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 494             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 495             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 496
 497             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 498             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 499             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 500             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 501
 502             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 503             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 504             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 505             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 506
 507             /* Load parameters for j particles */
 508             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 509             vdwjidx0A        = 2*vdwtype[jnrA+0];
 510
 511             fjx0             = _fjsp_setzero_v2r8();
 512             fjy0             = _fjsp_setzero_v2r8();
 513             fjz0             = _fjsp_setzero_v2r8();
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 520             {
 521
 522             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 523
 524             /* Compute parameters for interactions between i and j atoms */
 525             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 526
 527             c6grid_00       = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
 528
 529             /* Analytical LJ-PME */
 530             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 531             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 532             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 533             exponent         = gmx_simd_exp_d(-ewcljrsq);
 534             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 535             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 536             /* vvdw6 = [C6 - C6grid * (1-poly)]/r6 */
 537             vvdw6            = _fjsp_mul_v2r8(_fjsp_madd_v2r8(-c6grid_00,_fjsp_sub_v2r8(one,poly),c6_00),rinvsix);
 538             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 539             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 540                                _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));
 541             /* fvdw = vvdw12/r - (vvdw6/r + (C6grid * exponent * beta^6)/r) */
 542             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);
 543
 544             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 545
 546             /* Update potential sum for this i atom from the interaction with this j atom. */
 547             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 548             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 549             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 550
 551             fscal            = fvdw;
 552
 553             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 554
 555             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 556
 557             /* Update vectorial force */
 558             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 559             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 560             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 561
 562             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 563             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 564             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 565
 566             }
 567
 568             /**************************
 569              * CALCULATE INTERACTIONS *
 570              **************************/
 571
 572             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 573             {
 574
 575             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 576
 577             /* Compute parameters for interactions between i and j atoms */
 578             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 579
 580             /* EWALD ELECTROSTATICS */
 581
 582             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 583             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 584             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 585             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 586             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 587
 588             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 589             ewtabD           = _fjsp_setzero_v2r8();
 590             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 591             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 592             ewtabFn          = _fjsp_setzero_v2r8();
 593             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 594             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 595             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 596             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 597             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 598
 599             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 600
 601             /* Update potential sum for this i atom from the interaction with this j atom. */
 602             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 603             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 604             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 605
 606             fscal            = felec;
 607
 608             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 609
 610             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 611
 612             /* Update vectorial force */
 613             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 614             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 615             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 616
 617             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 618             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 619             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 620
 621             }
 622
 623             /**************************
 624              * CALCULATE INTERACTIONS *
 625              **************************/
 626
 627             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 628             {
 629
 630             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 631
 632             /* Compute parameters for interactions between i and j atoms */
 633             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 634
 635             /* EWALD ELECTROSTATICS */
 636
 637             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 638             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 639             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 640             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 641             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 642
 643             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 644             ewtabD           = _fjsp_setzero_v2r8();
 645             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 646             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 647             ewtabFn          = _fjsp_setzero_v2r8();
 648             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 649             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 650             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 651             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 652             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 653
 654             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 655
 656             /* Update potential sum for this i atom from the interaction with this j atom. */
 657             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 658             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 659             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 660
 661             fscal            = felec;
 662
 663             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 664
 665             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 666
 667             /* Update vectorial force */
 668             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 669             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 670             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 671
 672             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 673             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 674             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 675
 676             }
 677
 678             /**************************
 679              * CALCULATE INTERACTIONS *
 680              **************************/
 681
 682             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 683             {
 684
 685             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 686
 687             /* Compute parameters for interactions between i and j atoms */
 688             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 689
 690             /* EWALD ELECTROSTATICS */
 691
 692             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 693             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 694             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 695             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 696             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 697
 698             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 699             ewtabD           = _fjsp_setzero_v2r8();
 700             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 701             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 702             ewtabFn          = _fjsp_setzero_v2r8();
 703             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 704             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 705             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 706             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv30,sh_ewald),velec));
 707             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 708
 709             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 710
 711             /* Update potential sum for this i atom from the interaction with this j atom. */
 712             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 713             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 714             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 715
 716             fscal            = felec;
 717
 718             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 719
 720             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 721
 722             /* Update vectorial force */
 723             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 724             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 725             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 726
 727             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 728             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 729             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 730
 731             }
 732
 733             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 734
 735             /* Inner loop uses 209 flops */
 736         }
 737
 738         /* End of innermost loop */
 739
 740         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 741                                               f+i_coord_offset,fshift+i_shift_offset);
 742
 743         ggid                        = gid[iidx];
 744         /* Update potential energies */
 745         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 746         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 747
 748         /* Increment number of inner iterations */
 749         inneriter                  += j_index_end - j_index_start;
 750
 751         /* Outer loop uses 26 flops */
 752     }
 753
 754     /* Increment number of outer iterations */
 755     outeriter        += nri;
 756
 757     /* Update outer/inner flops */
 758
 759     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*209);
 760 }
 761 /*
 762  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_sparc64_hpc_ace_double
 763  * Electrostatics interaction: Ewald
 764  * VdW interaction:            LJEwald
 765  * Geometry:                   Water4-Particle
 766  * Calculate force/pot:        Force
 767  */
 768 void
 769 nb_kernel_ElecEwSh_VdwLJEwSh_GeomW4P1_F_sparc64_hpc_ace_double
 770                     (t_nblist                    * gmx_restrict       nlist,
 771                      rvec                        * gmx_restrict          xx,
 772                      rvec                        * gmx_restrict          ff,
 773                      t_forcerec                  * gmx_restrict          fr,
 774                      t_mdatoms                   * gmx_restrict     mdatoms,
 775                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 776                      t_nrnb                      * gmx_restrict        nrnb)
 777 {
 778     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 779      * just 0 for non-waters.
 780      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 781      * jnr indices corresponding to data put in the four positions in the SIMD register.
 782      */
 783     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 784     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 785     int              jnrA,jnrB;
 786     int              j_coord_offsetA,j_coord_offsetB;
 787     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 788     real             rcutoff_scalar;
 789     real             *shiftvec,*fshift,*x,*f;
 790     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 791     int              vdwioffset0;
 792     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 793     int              vdwioffset1;
 794     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 795     int              vdwioffset2;
 796     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 797     int              vdwioffset3;
 798     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 799     int              vdwjidx0A,vdwjidx0B;
 800     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 801     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 802     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 803     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 804     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 805     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 806     real             *charge;
 807     int              nvdwtype;
 808     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 809     int              *vdwtype;
 810     real             *vdwparam;
 811     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 812     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 813     _fjsp_v2r8           c6grid_00;
 814     _fjsp_v2r8           c6grid_10;
 815     _fjsp_v2r8           c6grid_20;
 816     _fjsp_v2r8           c6grid_30;
 817     real                 *vdwgridparam;
 818     _fjsp_v2r8           ewclj,ewclj2,ewclj6,ewcljrsq,poly,exponent,f6A,f6B,sh_lj_ewald;
 819     _fjsp_v2r8           one_half = gmx_fjsp_set1_v2r8(0.5);
 820     _fjsp_v2r8           minus_one = gmx_fjsp_set1_v2r8(-1.0);
 821     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 822     real             *ewtab;
 823     _fjsp_v2r8       itab_tmp;
 824     _fjsp_v2r8       dummy_mask,cutoff_mask;
 825     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 826     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 827     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 828
 829     x                = xx[0];
 830     f                = ff[0];
 831
 832     nri              = nlist->nri;
 833     iinr             = nlist->iinr;
 834     jindex           = nlist->jindex;
 835     jjnr             = nlist->jjnr;
 836     shiftidx         = nlist->shift;
 837     gid              = nlist->gid;
 838     shiftvec         = fr->shift_vec[0];
 839     fshift           = fr->fshift[0];
 840     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 841     charge           = mdatoms->chargeA;
 842     nvdwtype         = fr->ntype;
 843     vdwparam         = fr->nbfp;
 844     vdwtype          = mdatoms->typeA;
 845     vdwgridparam     = fr->ljpme_c6grid;
 846     sh_lj_ewald      = gmx_fjsp_set1_v2r8(fr->ic->sh_lj_ewald);
 847     ewclj            = gmx_fjsp_set1_v2r8(fr->ewaldcoeff_lj);
 848     ewclj2           = _fjsp_mul_v2r8(minus_one,_fjsp_mul_v2r8(ewclj,ewclj));
 849
 850     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 851     ewtab            = fr->ic->tabq_coul_F;
 852     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 853     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 854
 855     /* Setup water-specific parameters */
 856     inr              = nlist->iinr[0];
 857     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 858     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 859     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 860     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 861
 862     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 863     rcutoff_scalar   = fr->rcoulomb;
 864     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 865     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 866
 867     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 868     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 869
 870     /* Avoid stupid compiler warnings */
 871     jnrA = jnrB = 0;
 872     j_coord_offsetA = 0;
 873     j_coord_offsetB = 0;
 874
 875     outeriter        = 0;
 876     inneriter        = 0;
 877
 878     /* Start outer loop over neighborlists */
 879     for(iidx=0; iidx<nri; iidx++)
 880     {
 881         /* Load shift vector for this list */
 882         i_shift_offset   = DIM*shiftidx[iidx];
 883
 884         /* Load limits for loop over neighbors */
 885         j_index_start    = jindex[iidx];
 886         j_index_end      = jindex[iidx+1];
 887
 888         /* Get outer coordinate index */
 889         inr              = iinr[iidx];
 890         i_coord_offset   = DIM*inr;
 891
 892         /* Load i particle coords and add shift vector */
 893         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 894                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 895
 896         fix0             = _fjsp_setzero_v2r8();
 897         fiy0             = _fjsp_setzero_v2r8();
 898         fiz0             = _fjsp_setzero_v2r8();
 899         fix1             = _fjsp_setzero_v2r8();
 900         fiy1             = _fjsp_setzero_v2r8();
 901         fiz1             = _fjsp_setzero_v2r8();
 902         fix2             = _fjsp_setzero_v2r8();
 903         fiy2             = _fjsp_setzero_v2r8();
 904         fiz2             = _fjsp_setzero_v2r8();
 905         fix3             = _fjsp_setzero_v2r8();
 906         fiy3             = _fjsp_setzero_v2r8();
 907         fiz3             = _fjsp_setzero_v2r8();
 908
 909         /* Start inner kernel loop */
 910         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 911         {
 912
 913             /* Get j neighbor index, and coordinate index */
 914             jnrA             = jjnr[jidx];
 915             jnrB             = jjnr[jidx+1];
 916             j_coord_offsetA  = DIM*jnrA;
 917             j_coord_offsetB  = DIM*jnrB;
 918
 919             /* load j atom coordinates */
 920             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 921                                               &jx0,&jy0,&jz0);
 922
 923             /* Calculate displacement vector */
 924             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 925             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 926             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 927             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 928             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 929             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 930             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 931             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 932             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 933             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 934             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 935             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 936
 937             /* Calculate squared distance and things based on it */
 938             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 939             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 940             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 941             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 942
 943             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 944             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 945             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 946             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 947
 948             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 949             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 950             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 951             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 952
 953             /* Load parameters for j particles */
 954             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 955             vdwjidx0A        = 2*vdwtype[jnrA+0];
 956             vdwjidx0B        = 2*vdwtype[jnrB+0];
 957
 958             fjx0             = _fjsp_setzero_v2r8();
 959             fjy0             = _fjsp_setzero_v2r8();
 960             fjz0             = _fjsp_setzero_v2r8();
 961
 962             /**************************
 963              * CALCULATE INTERACTIONS *
 964              **************************/
 965
 966             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 967             {
 968
 969             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 970
 971             /* Compute parameters for interactions between i and j atoms */
 972             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 973                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 974
 975             c6grid_00       = gmx_fjsp_load_2real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A,
 976                                                                    vdwgridparam+vdwioffset0+vdwjidx0B);
 977
 978             /* Analytical LJ-PME */
 979             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 980             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
 981             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
 982             exponent         = gmx_simd_exp_d(-ewcljrsq);
 983             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
 984             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
 985             /* f6A = 6 * C6grid * (1 - poly) */
 986             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
 987             /* f6B = C6grid * exponent * beta^6 */
 988             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
 989             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
 990             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
 991
 992             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 993
 994             fscal            = fvdw;
 995
 996             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 997
 998             /* Update vectorial force */
 999             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1000             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1001             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1002
1003             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1004             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1005             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1006
1007             }
1008
1009             /**************************
1010              * CALCULATE INTERACTIONS *
1011              **************************/
1012
1013             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1014             {
1015
1016             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1017
1018             /* Compute parameters for interactions between i and j atoms */
1019             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1020
1021             /* EWALD ELECTROSTATICS */
1022
1023             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1024             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1025             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1026             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1027             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1028
1029             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1030                                          &ewtabF,&ewtabFn);
1031             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1032             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1033
1034             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1035
1036             fscal            = felec;
1037
1038             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1039
1040             /* Update vectorial force */
1041             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1042             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1043             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1044
1045             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1046             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1047             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1048
1049             }
1050
1051             /**************************
1052              * CALCULATE INTERACTIONS *
1053              **************************/
1054
1055             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1056             {
1057
1058             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1059
1060             /* Compute parameters for interactions between i and j atoms */
1061             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1062
1063             /* EWALD ELECTROSTATICS */
1064
1065             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1066             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1067             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1068             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1069             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1070
1071             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1072                                          &ewtabF,&ewtabFn);
1073             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1074             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1075
1076             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1077
1078             fscal            = felec;
1079
1080             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1081
1082             /* Update vectorial force */
1083             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1084             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1085             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1086
1087             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1088             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1089             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1090
1091             }
1092
1093             /**************************
1094              * CALCULATE INTERACTIONS *
1095              **************************/
1096
1097             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1098             {
1099
1100             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1101
1102             /* Compute parameters for interactions between i and j atoms */
1103             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1104
1105             /* EWALD ELECTROSTATICS */
1106
1107             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1108             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1109             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1110             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1111             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1112
1113             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
1114                                          &ewtabF,&ewtabFn);
1115             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1116             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1117
1118             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1119
1120             fscal            = felec;
1121
1122             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1123
1124             /* Update vectorial force */
1125             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1126             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1127             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1128
1129             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1130             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1131             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1132
1133             }
1134
1135             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
1136
1137             /* Inner loop uses 180 flops */
1138         }
1139
1140         if(jidx<j_index_end)
1141         {
1142
1143             jnrA             = jjnr[jidx];
1144             j_coord_offsetA  = DIM*jnrA;
1145
1146             /* load j atom coordinates */
1147             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1148                                               &jx0,&jy0,&jz0);
1149
1150             /* Calculate displacement vector */
1151             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1152             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1153             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1154             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1155             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1156             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1157             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1158             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1159             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1160             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1161             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1162             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1163
1164             /* Calculate squared distance and things based on it */
1165             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1166             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1167             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1168             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1169
1170             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1171             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1172             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1173             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1174
1175             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1176             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1177             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1178             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1179
1180             /* Load parameters for j particles */
1181             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1182             vdwjidx0A        = 2*vdwtype[jnrA+0];
1183
1184             fjx0             = _fjsp_setzero_v2r8();
1185             fjy0             = _fjsp_setzero_v2r8();
1186             fjz0             = _fjsp_setzero_v2r8();
1187
1188             /**************************
1189              * CALCULATE INTERACTIONS *
1190              **************************/
1191
1192             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1193             {
1194
1195             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1196
1197             /* Compute parameters for interactions between i and j atoms */
1198             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1199
1200             c6grid_00       = gmx_fjsp_load_1real_swizzle_v2r8(vdwgridparam+vdwioffset0+vdwjidx0A);
1201
1202             /* Analytical LJ-PME */
1203             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1204             ewcljrsq         = _fjsp_mul_v2r8(ewclj2,rsq00);
1205             ewclj6           = _fjsp_mul_v2r8(ewclj2,_fjsp_mul_v2r8(ewclj2,ewclj2));
1206             exponent         = gmx_simd_exp_d(-ewcljrsq);
1207             /* poly = exp(-(beta*r)^2) * (1 + (beta*r)^2 + (beta*r)^4 /2) */
1208             poly             = _fjsp_mul_v2r8(exponent,_fjsp_madd_v2r8(_fjsp_mul_v2r8(ewcljrsq,ewcljrsq),one_half,_fjsp_sub_v2r8(one,ewcljrsq)));
1209             /* f6A = 6 * C6grid * (1 - poly) */
1210             f6A              = _fjsp_mul_v2r8(c6grid_00,_fjsp_msub_v2r8(one,poly));
1211             /* f6B = C6grid * exponent * beta^6 */
1212             f6B              = _fjsp_mul_v2r8(_fjsp_mul_v2r8(c6grid_00,one_sixth),_fjsp_mul_v2r8(exponent,ewclj6));
1213             /* fvdw = 12*C12/r13 - ((6*C6 - f6A)/r6 + f6B)/r */
1214             fvdw              = _fjsp_mul_v2r8(_fjsp_madd_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,_fjsp_sub_v2r8(c6_00,f6A)),rinvsix,f6B),rinvsq00);
1215
1216             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1217
1218             fscal            = fvdw;
1219
1220             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1221
1222             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1223
1224             /* Update vectorial force */
1225             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1226             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1227             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1228
1229             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1230             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1231             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1232
1233             }
1234
1235             /**************************
1236              * CALCULATE INTERACTIONS *
1237              **************************/
1238
1239             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1240             {
1241
1242             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1243
1244             /* Compute parameters for interactions between i and j atoms */
1245             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1246
1247             /* EWALD ELECTROSTATICS */
1248
1249             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1250             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1251             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1252             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1253             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1254
1255             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1256             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1257             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1258
1259             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1260
1261             fscal            = felec;
1262
1263             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1264
1265             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1266
1267             /* Update vectorial force */
1268             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1269             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1270             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1271
1272             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1273             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1274             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1275
1276             }
1277
1278             /**************************
1279              * CALCULATE INTERACTIONS *
1280              **************************/
1281
1282             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1283             {
1284
1285             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1286
1287             /* Compute parameters for interactions between i and j atoms */
1288             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1289
1290             /* EWALD ELECTROSTATICS */
1291
1292             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1293             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1294             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1295             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1296             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1297
1298             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1299             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1300             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1301
1302             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1303
1304             fscal            = felec;
1305
1306             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1307
1308             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1309
1310             /* Update vectorial force */
1311             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1312             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1313             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1314
1315             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1316             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1317             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1318
1319             }
1320
1321             /**************************
1322              * CALCULATE INTERACTIONS *
1323              **************************/
1324
1325             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1326             {
1327
1328             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1329
1330             /* Compute parameters for interactions between i and j atoms */
1331             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1332
1333             /* EWALD ELECTROSTATICS */
1334
1335             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1336             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1337             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1338             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1339             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1340
1341             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1342             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1343             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1344
1345             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1346
1347             fscal            = felec;
1348
1349             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1350
1351             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1352
1353             /* Update vectorial force */
1354             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1355             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1356             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1357
1358             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1359             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1360             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1361
1362             }
1363
1364             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1365
1366             /* Inner loop uses 180 flops */
1367         }
1368
1369         /* End of innermost loop */
1370
1371         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1372                                               f+i_coord_offset,fshift+i_shift_offset);
1373
1374         /* Increment number of inner iterations */
1375         inneriter                  += j_index_end - j_index_start;
1376
1377         /* Outer loop uses 24 flops */
1378     }
1379
1380     /* Increment number of outer iterations */
1381     outeriter        += nri;
1382
1383     /* Update outer/inner flops */
1384
1385     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*180);
1386 }