src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  99     real             *ewtab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 124     ewtab            = fr->ic->tabq_coul_FDV0;
 125     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 126     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 136     rcutoff_scalar   = fr->rcoulomb;
 137     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 138     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 139
 140     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 141     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 142
 143     /* Avoid stupid compiler warnings */
 144     jnrA = jnrB = 0;
 145     j_coord_offsetA = 0;
 146     j_coord_offsetB = 0;
 147
 148     outeriter        = 0;
 149     inneriter        = 0;
 150
 151     /* Start outer loop over neighborlists */
 152     for(iidx=0; iidx<nri; iidx++)
 153     {
 154         /* Load shift vector for this list */
 155         i_shift_offset   = DIM*shiftidx[iidx];
 156
 157         /* Load limits for loop over neighbors */
 158         j_index_start    = jindex[iidx];
 159         j_index_end      = jindex[iidx+1];
 160
 161         /* Get outer coordinate index */
 162         inr              = iinr[iidx];
 163         i_coord_offset   = DIM*inr;
 164
 165         /* Load i particle coords and add shift vector */
 166         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 167                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 168
 169         fix0             = _fjsp_setzero_v2r8();
 170         fiy0             = _fjsp_setzero_v2r8();
 171         fiz0             = _fjsp_setzero_v2r8();
 172         fix1             = _fjsp_setzero_v2r8();
 173         fiy1             = _fjsp_setzero_v2r8();
 174         fiz1             = _fjsp_setzero_v2r8();
 175         fix2             = _fjsp_setzero_v2r8();
 176         fiy2             = _fjsp_setzero_v2r8();
 177         fiz2             = _fjsp_setzero_v2r8();
 178
 179         /* Reset potential sums */
 180         velecsum         = _fjsp_setzero_v2r8();
 181         vvdwsum          = _fjsp_setzero_v2r8();
 182
 183         /* Start inner kernel loop */
 184         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 185         {
 186
 187             /* Get j neighbor index, and coordinate index */
 188             jnrA             = jjnr[jidx];
 189             jnrB             = jjnr[jidx+1];
 190             j_coord_offsetA  = DIM*jnrA;
 191             j_coord_offsetB  = DIM*jnrB;
 192
 193             /* load j atom coordinates */
 194             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 195                                               &jx0,&jy0,&jz0);
 196
 197             /* Calculate displacement vector */
 198             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 199             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 200             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 201             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 202             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 203             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 204             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 205             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 206             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 207
 208             /* Calculate squared distance and things based on it */
 209             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 210             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 211             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 212
 213             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 214             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 215             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 216
 217             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 218             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 219             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 220
 221             /* Load parameters for j particles */
 222             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 223             vdwjidx0A        = 2*vdwtype[jnrA+0];
 224             vdwjidx0B        = 2*vdwtype[jnrB+0];
 225
 226             fjx0             = _fjsp_setzero_v2r8();
 227             fjy0             = _fjsp_setzero_v2r8();
 228             fjz0             = _fjsp_setzero_v2r8();
 229
 230             /**************************
 231              * CALCULATE INTERACTIONS *
 232              **************************/
 233
 234             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 235             {
 236
 237             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 238
 239             /* Compute parameters for interactions between i and j atoms */
 240             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 241             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 242                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 243
 244             /* EWALD ELECTROSTATICS */
 245
 246             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 247             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 248             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 249             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 250             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 251
 252             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 253             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 254             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 255             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 256             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 257             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 258             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 259             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 260             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 261             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 262
 263             /* LENNARD-JONES DISPERSION/REPULSION */
 264
 265             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 266             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 267             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 268             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 269                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 270             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 271
 272             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 273
 274             /* Update potential sum for this i atom from the interaction with this j atom. */
 275             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 276             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 277             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 278             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 279
 280             fscal            = _fjsp_add_v2r8(felec,fvdw);
 281
 282             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 283
 284             /* Update vectorial force */
 285             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 286             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 287             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 288
 289             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 290             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 291             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 292
 293             }
 294
 295             /**************************
 296              * CALCULATE INTERACTIONS *
 297              **************************/
 298
 299             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 300             {
 301
 302             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 303
 304             /* Compute parameters for interactions between i and j atoms */
 305             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 306
 307             /* EWALD ELECTROSTATICS */
 308
 309             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 310             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 311             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 312             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 313             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 314
 315             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 316             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 317             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 318             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 319             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 320             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 321             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 322             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 323             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 324             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 325
 326             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 327
 328             /* Update potential sum for this i atom from the interaction with this j atom. */
 329             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 335
 336             /* Update vectorial force */
 337             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 338             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 339             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 340
 341             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 342             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 343             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 344
 345             }
 346
 347             /**************************
 348              * CALCULATE INTERACTIONS *
 349              **************************/
 350
 351             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 352             {
 353
 354             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 355
 356             /* Compute parameters for interactions between i and j atoms */
 357             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 358
 359             /* EWALD ELECTROSTATICS */
 360
 361             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 362             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 363             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 364             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 365             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 366
 367             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 368             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 369             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 370             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 371             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 372             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 373             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 374             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 375             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 376             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 377
 378             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 379
 380             /* Update potential sum for this i atom from the interaction with this j atom. */
 381             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 382             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 383
 384             fscal            = felec;
 385
 386             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 387
 388             /* Update vectorial force */
 389             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 390             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 391             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 392
 393             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 394             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 395             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 396
 397             }
 398
 399             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 400
 401             /* Inner loop uses 168 flops */
 402         }
 403
 404         if(jidx<j_index_end)
 405         {
 406
 407             jnrA             = jjnr[jidx];
 408             j_coord_offsetA  = DIM*jnrA;
 409
 410             /* load j atom coordinates */
 411             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 412                                               &jx0,&jy0,&jz0);
 413
 414             /* Calculate displacement vector */
 415             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 416             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 417             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 418             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 419             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 420             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 421             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 422             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 423             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 424
 425             /* Calculate squared distance and things based on it */
 426             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 427             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 428             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 429
 430             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 431             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 432             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 433
 434             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 435             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 436             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 437
 438             /* Load parameters for j particles */
 439             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 440             vdwjidx0A        = 2*vdwtype[jnrA+0];
 441
 442             fjx0             = _fjsp_setzero_v2r8();
 443             fjy0             = _fjsp_setzero_v2r8();
 444             fjz0             = _fjsp_setzero_v2r8();
 445
 446             /**************************
 447              * CALCULATE INTERACTIONS *
 448              **************************/
 449
 450             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 451             {
 452
 453             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 454
 455             /* Compute parameters for interactions between i and j atoms */
 456             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 457             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 458
 459             /* EWALD ELECTROSTATICS */
 460
 461             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 462             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 463             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 464             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 465             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 466
 467             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 468             ewtabD           = _fjsp_setzero_v2r8();
 469             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 470             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 471             ewtabFn          = _fjsp_setzero_v2r8();
 472             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 473             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 474             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 475             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 476             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 477
 478             /* LENNARD-JONES DISPERSION/REPULSION */
 479
 480             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 481             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 482             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 483             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 484                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 485             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 486
 487             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 488
 489             /* Update potential sum for this i atom from the interaction with this j atom. */
 490             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 491             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 492             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 493             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 494             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 495             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 496
 497             fscal            = _fjsp_add_v2r8(felec,fvdw);
 498
 499             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 500
 501             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 502
 503             /* Update vectorial force */
 504             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 505             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 506             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 507
 508             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 509             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 510             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 511
 512             }
 513
 514             /**************************
 515              * CALCULATE INTERACTIONS *
 516              **************************/
 517
 518             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 519             {
 520
 521             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 522
 523             /* Compute parameters for interactions between i and j atoms */
 524             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 525
 526             /* EWALD ELECTROSTATICS */
 527
 528             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 529             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 530             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 531             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 532             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 533
 534             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 535             ewtabD           = _fjsp_setzero_v2r8();
 536             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 537             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 538             ewtabFn          = _fjsp_setzero_v2r8();
 539             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 540             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 541             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 542             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 543             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 544
 545             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 546
 547             /* Update potential sum for this i atom from the interaction with this j atom. */
 548             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 549             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 550             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 551
 552             fscal            = felec;
 553
 554             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 555
 556             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 557
 558             /* Update vectorial force */
 559             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 560             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 561             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 562
 563             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 564             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 565             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 566
 567             }
 568
 569             /**************************
 570              * CALCULATE INTERACTIONS *
 571              **************************/
 572
 573             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 574             {
 575
 576             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 577
 578             /* Compute parameters for interactions between i and j atoms */
 579             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 580
 581             /* EWALD ELECTROSTATICS */
 582
 583             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 584             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 585             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 586             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 587             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 588
 589             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 590             ewtabD           = _fjsp_setzero_v2r8();
 591             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 592             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 593             ewtabFn          = _fjsp_setzero_v2r8();
 594             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 595             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 596             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 597             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 598             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 599
 600             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 601
 602             /* Update potential sum for this i atom from the interaction with this j atom. */
 603             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 604             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 605             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 606
 607             fscal            = felec;
 608
 609             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 610
 611             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 612
 613             /* Update vectorial force */
 614             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 615             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 616             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 617
 618             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 619             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 620             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 621
 622             }
 623
 624             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 625
 626             /* Inner loop uses 168 flops */
 627         }
 628
 629         /* End of innermost loop */
 630
 631         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 632                                               f+i_coord_offset,fshift+i_shift_offset);
 633
 634         ggid                        = gid[iidx];
 635         /* Update potential energies */
 636         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 637         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 638
 639         /* Increment number of inner iterations */
 640         inneriter                  += j_index_end - j_index_start;
 641
 642         /* Outer loop uses 20 flops */
 643     }
 644
 645     /* Increment number of outer iterations */
 646     outeriter        += nri;
 647
 648     /* Update outer/inner flops */
 649
 650     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*168);
 651 }
 652 /*
 653  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_F_sparc64_hpc_ace_double
 654  * Electrostatics interaction: Ewald
 655  * VdW interaction:            LennardJones
 656  * Geometry:                   Water3-Particle
 657  * Calculate force/pot:        Force
 658  */
 659 void
 660 nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_F_sparc64_hpc_ace_double
 661                     (t_nblist                    * gmx_restrict       nlist,
 662                      rvec                        * gmx_restrict          xx,
 663                      rvec                        * gmx_restrict          ff,
 664                      t_forcerec                  * gmx_restrict          fr,
 665                      t_mdatoms                   * gmx_restrict     mdatoms,
 666                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 667                      t_nrnb                      * gmx_restrict        nrnb)
 668 {
 669     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 670      * just 0 for non-waters.
 671      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 672      * jnr indices corresponding to data put in the four positions in the SIMD register.
 673      */
 674     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 675     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 676     int              jnrA,jnrB;
 677     int              j_coord_offsetA,j_coord_offsetB;
 678     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 679     real             rcutoff_scalar;
 680     real             *shiftvec,*fshift,*x,*f;
 681     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 682     int              vdwioffset0;
 683     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 684     int              vdwioffset1;
 685     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 686     int              vdwioffset2;
 687     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 688     int              vdwjidx0A,vdwjidx0B;
 689     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 690     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 691     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 692     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 693     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 694     real             *charge;
 695     int              nvdwtype;
 696     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 697     int              *vdwtype;
 698     real             *vdwparam;
 699     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 700     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 701     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 702     real             *ewtab;
 703     _fjsp_v2r8       itab_tmp;
 704     _fjsp_v2r8       dummy_mask,cutoff_mask;
 705     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 706     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 707     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 708
 709     x                = xx[0];
 710     f                = ff[0];
 711
 712     nri              = nlist->nri;
 713     iinr             = nlist->iinr;
 714     jindex           = nlist->jindex;
 715     jjnr             = nlist->jjnr;
 716     shiftidx         = nlist->shift;
 717     gid              = nlist->gid;
 718     shiftvec         = fr->shift_vec[0];
 719     fshift           = fr->fshift[0];
 720     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 721     charge           = mdatoms->chargeA;
 722     nvdwtype         = fr->ntype;
 723     vdwparam         = fr->nbfp;
 724     vdwtype          = mdatoms->typeA;
 725
 726     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 727     ewtab            = fr->ic->tabq_coul_F;
 728     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 729     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 730
 731     /* Setup water-specific parameters */
 732     inr              = nlist->iinr[0];
 733     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 734     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 735     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 736     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 737
 738     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 739     rcutoff_scalar   = fr->rcoulomb;
 740     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 741     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 742
 743     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 744     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 745
 746     /* Avoid stupid compiler warnings */
 747     jnrA = jnrB = 0;
 748     j_coord_offsetA = 0;
 749     j_coord_offsetB = 0;
 750
 751     outeriter        = 0;
 752     inneriter        = 0;
 753
 754     /* Start outer loop over neighborlists */
 755     for(iidx=0; iidx<nri; iidx++)
 756     {
 757         /* Load shift vector for this list */
 758         i_shift_offset   = DIM*shiftidx[iidx];
 759
 760         /* Load limits for loop over neighbors */
 761         j_index_start    = jindex[iidx];
 762         j_index_end      = jindex[iidx+1];
 763
 764         /* Get outer coordinate index */
 765         inr              = iinr[iidx];
 766         i_coord_offset   = DIM*inr;
 767
 768         /* Load i particle coords and add shift vector */
 769         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 770                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 771
 772         fix0             = _fjsp_setzero_v2r8();
 773         fiy0             = _fjsp_setzero_v2r8();
 774         fiz0             = _fjsp_setzero_v2r8();
 775         fix1             = _fjsp_setzero_v2r8();
 776         fiy1             = _fjsp_setzero_v2r8();
 777         fiz1             = _fjsp_setzero_v2r8();
 778         fix2             = _fjsp_setzero_v2r8();
 779         fiy2             = _fjsp_setzero_v2r8();
 780         fiz2             = _fjsp_setzero_v2r8();
 781
 782         /* Start inner kernel loop */
 783         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 784         {
 785
 786             /* Get j neighbor index, and coordinate index */
 787             jnrA             = jjnr[jidx];
 788             jnrB             = jjnr[jidx+1];
 789             j_coord_offsetA  = DIM*jnrA;
 790             j_coord_offsetB  = DIM*jnrB;
 791
 792             /* load j atom coordinates */
 793             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 794                                               &jx0,&jy0,&jz0);
 795
 796             /* Calculate displacement vector */
 797             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 798             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 799             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 800             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 801             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 802             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 803             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 804             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 805             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 806
 807             /* Calculate squared distance and things based on it */
 808             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 809             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 810             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 811
 812             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 813             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 814             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 815
 816             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 817             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 818             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 819
 820             /* Load parameters for j particles */
 821             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 822             vdwjidx0A        = 2*vdwtype[jnrA+0];
 823             vdwjidx0B        = 2*vdwtype[jnrB+0];
 824
 825             fjx0             = _fjsp_setzero_v2r8();
 826             fjy0             = _fjsp_setzero_v2r8();
 827             fjz0             = _fjsp_setzero_v2r8();
 828
 829             /**************************
 830              * CALCULATE INTERACTIONS *
 831              **************************/
 832
 833             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 834             {
 835
 836             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 837
 838             /* Compute parameters for interactions between i and j atoms */
 839             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 840             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 841                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 842
 843             /* EWALD ELECTROSTATICS */
 844
 845             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 846             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 847             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 848             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 849             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 850
 851             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 852                                          &ewtabF,&ewtabFn);
 853             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 854             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 855
 856             /* LENNARD-JONES DISPERSION/REPULSION */
 857
 858             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 859             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 860
 861             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 862
 863             fscal            = _fjsp_add_v2r8(felec,fvdw);
 864
 865             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 866
 867             /* Update vectorial force */
 868             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 869             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 870             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 871
 872             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 873             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 874             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 875
 876             }
 877
 878             /**************************
 879              * CALCULATE INTERACTIONS *
 880              **************************/
 881
 882             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 883             {
 884
 885             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 886
 887             /* Compute parameters for interactions between i and j atoms */
 888             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 889
 890             /* EWALD ELECTROSTATICS */
 891
 892             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 893             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 894             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 895             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 896             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 897
 898             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 899                                          &ewtabF,&ewtabFn);
 900             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 901             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 902
 903             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 904
 905             fscal            = felec;
 906
 907             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 908
 909             /* Update vectorial force */
 910             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 911             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 912             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 913
 914             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 915             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 916             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 917
 918             }
 919
 920             /**************************
 921              * CALCULATE INTERACTIONS *
 922              **************************/
 923
 924             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 925             {
 926
 927             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 928
 929             /* Compute parameters for interactions between i and j atoms */
 930             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 931
 932             /* EWALD ELECTROSTATICS */
 933
 934             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 935             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 936             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 937             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 938             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 939
 940             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 941                                          &ewtabF,&ewtabFn);
 942             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 943             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 944
 945             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 946
 947             fscal            = felec;
 948
 949             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 950
 951             /* Update vectorial force */
 952             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 953             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 954             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 955
 956             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 957             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 958             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 959
 960             }
 961
 962             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 963
 964             /* Inner loop uses 136 flops */
 965         }
 966
 967         if(jidx<j_index_end)
 968         {
 969
 970             jnrA             = jjnr[jidx];
 971             j_coord_offsetA  = DIM*jnrA;
 972
 973             /* load j atom coordinates */
 974             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 975                                               &jx0,&jy0,&jz0);
 976
 977             /* Calculate displacement vector */
 978             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 979             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 980             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 981             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 982             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 983             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 984             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 985             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 986             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 987
 988             /* Calculate squared distance and things based on it */
 989             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 990             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 991             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 992
 993             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 994             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 995             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 996
 997             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 998             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 999             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1000
1001             /* Load parameters for j particles */
1002             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1003             vdwjidx0A        = 2*vdwtype[jnrA+0];
1004
1005             fjx0             = _fjsp_setzero_v2r8();
1006             fjy0             = _fjsp_setzero_v2r8();
1007             fjz0             = _fjsp_setzero_v2r8();
1008
1009             /**************************
1010              * CALCULATE INTERACTIONS *
1011              **************************/
1012
1013             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1014             {
1015
1016             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1017
1018             /* Compute parameters for interactions between i and j atoms */
1019             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1020             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1021
1022             /* EWALD ELECTROSTATICS */
1023
1024             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1025             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1026             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1027             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1028             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1029
1030             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1031             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1032             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1033
1034             /* LENNARD-JONES DISPERSION/REPULSION */
1035
1036             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1037             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1038
1039             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1040
1041             fscal            = _fjsp_add_v2r8(felec,fvdw);
1042
1043             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1044
1045             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1046
1047             /* Update vectorial force */
1048             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1049             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1050             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1051
1052             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1053             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1054             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1055
1056             }
1057
1058             /**************************
1059              * CALCULATE INTERACTIONS *
1060              **************************/
1061
1062             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1063             {
1064
1065             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1066
1067             /* Compute parameters for interactions between i and j atoms */
1068             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1069
1070             /* EWALD ELECTROSTATICS */
1071
1072             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1073             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1074             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1075             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1076             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1077
1078             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1079             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1080             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1081
1082             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1083
1084             fscal            = felec;
1085
1086             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1087
1088             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1089
1090             /* Update vectorial force */
1091             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1092             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1093             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1094
1095             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1096             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1097             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1098
1099             }
1100
1101             /**************************
1102              * CALCULATE INTERACTIONS *
1103              **************************/
1104
1105             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1106             {
1107
1108             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1109
1110             /* Compute parameters for interactions between i and j atoms */
1111             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1112
1113             /* EWALD ELECTROSTATICS */
1114
1115             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1116             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1117             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1118             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1119             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1120
1121             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1122             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1123             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1124
1125             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1126
1127             fscal            = felec;
1128
1129             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1130
1131             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1132
1133             /* Update vectorial force */
1134             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1135             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1136             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1137
1138             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1139             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1140             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1141
1142             }
1143
1144             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1145
1146             /* Inner loop uses 136 flops */
1147         }
1148
1149         /* End of innermost loop */
1150
1151         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1152                                               f+i_coord_offset,fshift+i_shift_offset);
1153
1154         /* Increment number of inner iterations */
1155         inneriter                  += j_index_end - j_index_start;
1156
1157         /* Outer loop uses 18 flops */
1158     }
1159
1160     /* Increment number of outer iterations */
1161     outeriter        += nri;
1162
1163     /* Update outer/inner flops */
1164
1165     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*136);
1166 }