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, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              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       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 101     real             *ewtab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 126     ewtab            = fr->ic->tabq_coul_FDV0;
 127     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 128     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 129
 130     /* Setup water-specific parameters */
 131     inr              = nlist->iinr[0];
 132     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 133     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 134     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 135     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 136
 137     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 138     rcutoff_scalar   = fr->rcoulomb;
 139     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 140     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 141
 142     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 143     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 144
 145     /* Avoid stupid compiler warnings */
 146     jnrA = jnrB = 0;
 147     j_coord_offsetA = 0;
 148     j_coord_offsetB = 0;
 149
 150     outeriter        = 0;
 151     inneriter        = 0;
 152
 153     /* Start outer loop over neighborlists */
 154     for(iidx=0; iidx<nri; iidx++)
 155     {
 156         /* Load shift vector for this list */
 157         i_shift_offset   = DIM*shiftidx[iidx];
 158
 159         /* Load limits for loop over neighbors */
 160         j_index_start    = jindex[iidx];
 161         j_index_end      = jindex[iidx+1];
 162
 163         /* Get outer coordinate index */
 164         inr              = iinr[iidx];
 165         i_coord_offset   = DIM*inr;
 166
 167         /* Load i particle coords and add shift vector */
 168         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 169                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 170
 171         fix0             = _fjsp_setzero_v2r8();
 172         fiy0             = _fjsp_setzero_v2r8();
 173         fiz0             = _fjsp_setzero_v2r8();
 174         fix1             = _fjsp_setzero_v2r8();
 175         fiy1             = _fjsp_setzero_v2r8();
 176         fiz1             = _fjsp_setzero_v2r8();
 177         fix2             = _fjsp_setzero_v2r8();
 178         fiy2             = _fjsp_setzero_v2r8();
 179         fiz2             = _fjsp_setzero_v2r8();
 180
 181         /* Reset potential sums */
 182         velecsum         = _fjsp_setzero_v2r8();
 183         vvdwsum          = _fjsp_setzero_v2r8();
 184
 185         /* Start inner kernel loop */
 186         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 187         {
 188
 189             /* Get j neighbor index, and coordinate index */
 190             jnrA             = jjnr[jidx];
 191             jnrB             = jjnr[jidx+1];
 192             j_coord_offsetA  = DIM*jnrA;
 193             j_coord_offsetB  = DIM*jnrB;
 194
 195             /* load j atom coordinates */
 196             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 197                                               &jx0,&jy0,&jz0);
 198
 199             /* Calculate displacement vector */
 200             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 201             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 202             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 203             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 204             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 205             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 206             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 207             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 208             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 209
 210             /* Calculate squared distance and things based on it */
 211             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 212             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 213             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 214
 215             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 216             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 217             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 218
 219             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 220             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 221             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 222
 223             /* Load parameters for j particles */
 224             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 225             vdwjidx0A        = 2*vdwtype[jnrA+0];
 226             vdwjidx0B        = 2*vdwtype[jnrB+0];
 227
 228             fjx0             = _fjsp_setzero_v2r8();
 229             fjy0             = _fjsp_setzero_v2r8();
 230             fjz0             = _fjsp_setzero_v2r8();
 231
 232             /**************************
 233              * CALCULATE INTERACTIONS *
 234              **************************/
 235
 236             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 237             {
 238
 239             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 240
 241             /* Compute parameters for interactions between i and j atoms */
 242             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 243             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 244                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 245
 246             /* EWALD ELECTROSTATICS */
 247
 248             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 249             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 250             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 251             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 252             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 253
 254             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 255             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 256             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 257             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 258             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 259             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 260             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 261             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 262             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 263             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 264
 265             /* LENNARD-JONES DISPERSION/REPULSION */
 266
 267             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 268             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 269             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 270             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 271                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 272             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 273
 274             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 275
 276             /* Update potential sum for this i atom from the interaction with this j atom. */
 277             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 278             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 279             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 280             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 281
 282             fscal            = _fjsp_add_v2r8(felec,fvdw);
 283
 284             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 285
 286             /* Update vectorial force */
 287             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 288             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 289             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 290
 291             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 292             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 293             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 294
 295             }
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 302             {
 303
 304             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 305
 306             /* Compute parameters for interactions between i and j atoms */
 307             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 308
 309             /* EWALD ELECTROSTATICS */
 310
 311             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 312             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 313             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 314             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 315             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 316
 317             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 318             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 319             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 320             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 321             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 322             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 323             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 324             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 325             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 326             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 327
 328             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 329
 330             /* Update potential sum for this i atom from the interaction with this j atom. */
 331             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 332             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 333
 334             fscal            = felec;
 335
 336             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 337
 338             /* Update vectorial force */
 339             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 340             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 341             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 342
 343             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 344             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 345             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 346
 347             }
 348
 349             /**************************
 350              * CALCULATE INTERACTIONS *
 351              **************************/
 352
 353             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 354             {
 355
 356             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 357
 358             /* Compute parameters for interactions between i and j atoms */
 359             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 360
 361             /* EWALD ELECTROSTATICS */
 362
 363             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 364             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 365             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 366             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 367             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 368
 369             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 370             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 371             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 372             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 373             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 374             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 375             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 376             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 377             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 378             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 379
 380             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 384             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 385
 386             fscal            = felec;
 387
 388             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 389
 390             /* Update vectorial force */
 391             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 392             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 393             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 394
 395             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 396             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 397             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 398
 399             }
 400
 401             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 402
 403             /* Inner loop uses 168 flops */
 404         }
 405
 406         if(jidx<j_index_end)
 407         {
 408
 409             jnrA             = jjnr[jidx];
 410             j_coord_offsetA  = DIM*jnrA;
 411
 412             /* load j atom coordinates */
 413             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 414                                               &jx0,&jy0,&jz0);
 415
 416             /* Calculate displacement vector */
 417             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 418             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 419             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 420             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 421             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 422             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 423             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 424             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 425             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 426
 427             /* Calculate squared distance and things based on it */
 428             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 429             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 430             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 431
 432             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 433             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 434             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 435
 436             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 437             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 438             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 439
 440             /* Load parameters for j particles */
 441             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 442             vdwjidx0A        = 2*vdwtype[jnrA+0];
 443
 444             fjx0             = _fjsp_setzero_v2r8();
 445             fjy0             = _fjsp_setzero_v2r8();
 446             fjz0             = _fjsp_setzero_v2r8();
 447
 448             /**************************
 449              * CALCULATE INTERACTIONS *
 450              **************************/
 451
 452             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 453             {
 454
 455             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 456
 457             /* Compute parameters for interactions between i and j atoms */
 458             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 459             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 460
 461             /* EWALD ELECTROSTATICS */
 462
 463             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 464             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 465             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 466             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 467             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 468
 469             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 470             ewtabD           = _fjsp_setzero_v2r8();
 471             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 472             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 473             ewtabFn          = _fjsp_setzero_v2r8();
 474             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 475             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 476             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 477             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv00,sh_ewald),velec));
 478             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 479
 480             /* LENNARD-JONES DISPERSION/REPULSION */
 481
 482             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 483             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 484             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 485             vvdw             = _fjsp_msub_v2r8(_fjsp_nmsub_v2r8(c12_00,_fjsp_mul_v2r8(sh_vdw_invrcut6,sh_vdw_invrcut6),vvdw12),one_twelfth,
 486                                            _fjsp_mul_v2r8(_fjsp_nmsub_v2r8( c6_00,sh_vdw_invrcut6,vvdw6),one_sixth));
 487             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 488
 489             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 490
 491             /* Update potential sum for this i atom from the interaction with this j atom. */
 492             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 493             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 494             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 495             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 496             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 497             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 498
 499             fscal            = _fjsp_add_v2r8(felec,fvdw);
 500
 501             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 502
 503             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 504
 505             /* Update vectorial force */
 506             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 507             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 508             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 509
 510             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 511             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 512             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 513
 514             }
 515
 516             /**************************
 517              * CALCULATE INTERACTIONS *
 518              **************************/
 519
 520             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 521             {
 522
 523             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 524
 525             /* Compute parameters for interactions between i and j atoms */
 526             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 527
 528             /* EWALD ELECTROSTATICS */
 529
 530             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 531             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 532             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 533             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 534             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 535
 536             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 537             ewtabD           = _fjsp_setzero_v2r8();
 538             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 539             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 540             ewtabFn          = _fjsp_setzero_v2r8();
 541             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 542             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 543             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 544             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv10,sh_ewald),velec));
 545             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 546
 547             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 548
 549             /* Update potential sum for this i atom from the interaction with this j atom. */
 550             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 551             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 552             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 553
 554             fscal            = felec;
 555
 556             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 557
 558             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 559
 560             /* Update vectorial force */
 561             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 562             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 563             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 564
 565             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 566             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 567             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 568
 569             }
 570
 571             /**************************
 572              * CALCULATE INTERACTIONS *
 573              **************************/
 574
 575             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 576             {
 577
 578             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 579
 580             /* Compute parameters for interactions between i and j atoms */
 581             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 582
 583             /* EWALD ELECTROSTATICS */
 584
 585             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 586             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 587             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 588             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 589             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 590
 591             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 592             ewtabD           = _fjsp_setzero_v2r8();
 593             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 594             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 595             ewtabFn          = _fjsp_setzero_v2r8();
 596             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 597             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 598             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 599             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_sub_v2r8(rinv20,sh_ewald),velec));
 600             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 601
 602             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 603
 604             /* Update potential sum for this i atom from the interaction with this j atom. */
 605             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 606             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 607             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 608
 609             fscal            = felec;
 610
 611             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 612
 613             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 614
 615             /* Update vectorial force */
 616             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 617             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 618             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 619
 620             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 621             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 622             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 623
 624             }
 625
 626             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 627
 628             /* Inner loop uses 168 flops */
 629         }
 630
 631         /* End of innermost loop */
 632
 633         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 634                                               f+i_coord_offset,fshift+i_shift_offset);
 635
 636         ggid                        = gid[iidx];
 637         /* Update potential energies */
 638         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 639         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 640
 641         /* Increment number of inner iterations */
 642         inneriter                  += j_index_end - j_index_start;
 643
 644         /* Outer loop uses 20 flops */
 645     }
 646
 647     /* Increment number of outer iterations */
 648     outeriter        += nri;
 649
 650     /* Update outer/inner flops */
 651
 652     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*168);
 653 }
 654 /*
 655  * Gromacs nonbonded kernel:   nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_F_sparc64_hpc_ace_double
 656  * Electrostatics interaction: Ewald
 657  * VdW interaction:            LennardJones
 658  * Geometry:                   Water3-Particle
 659  * Calculate force/pot:        Force
 660  */
 661 void
 662 nb_kernel_ElecEwSh_VdwLJSh_GeomW3P1_F_sparc64_hpc_ace_double
 663                     (t_nblist                    * gmx_restrict       nlist,
 664                      rvec                        * gmx_restrict          xx,
 665                      rvec                        * gmx_restrict          ff,
 666                      t_forcerec                  * gmx_restrict          fr,
 667                      t_mdatoms                   * gmx_restrict     mdatoms,
 668                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 669                      t_nrnb                      * gmx_restrict        nrnb)
 670 {
 671     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 672      * just 0 for non-waters.
 673      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 674      * jnr indices corresponding to data put in the four positions in the SIMD register.
 675      */
 676     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 677     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 678     int              jnrA,jnrB;
 679     int              j_coord_offsetA,j_coord_offsetB;
 680     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 681     real             rcutoff_scalar;
 682     real             *shiftvec,*fshift,*x,*f;
 683     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 684     int              vdwioffset0;
 685     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 686     int              vdwioffset1;
 687     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 688     int              vdwioffset2;
 689     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 690     int              vdwjidx0A,vdwjidx0B;
 691     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 692     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 693     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 694     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 695     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 696     real             *charge;
 697     int              nvdwtype;
 698     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 699     int              *vdwtype;
 700     real             *vdwparam;
 701     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 702     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 703     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 704     real             *ewtab;
 705     _fjsp_v2r8       itab_tmp;
 706     _fjsp_v2r8       dummy_mask,cutoff_mask;
 707     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 708     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 709     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 710
 711     x                = xx[0];
 712     f                = ff[0];
 713
 714     nri              = nlist->nri;
 715     iinr             = nlist->iinr;
 716     jindex           = nlist->jindex;
 717     jjnr             = nlist->jjnr;
 718     shiftidx         = nlist->shift;
 719     gid              = nlist->gid;
 720     shiftvec         = fr->shift_vec[0];
 721     fshift           = fr->fshift[0];
 722     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 723     charge           = mdatoms->chargeA;
 724     nvdwtype         = fr->ntype;
 725     vdwparam         = fr->nbfp;
 726     vdwtype          = mdatoms->typeA;
 727
 728     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 729     ewtab            = fr->ic->tabq_coul_F;
 730     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 731     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 732
 733     /* Setup water-specific parameters */
 734     inr              = nlist->iinr[0];
 735     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 736     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 737     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 738     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 739
 740     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 741     rcutoff_scalar   = fr->rcoulomb;
 742     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 743     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 744
 745     sh_vdw_invrcut6  = gmx_fjsp_set1_v2r8(fr->ic->sh_invrc6);
 746     rvdw             = gmx_fjsp_set1_v2r8(fr->rvdw);
 747
 748     /* Avoid stupid compiler warnings */
 749     jnrA = jnrB = 0;
 750     j_coord_offsetA = 0;
 751     j_coord_offsetB = 0;
 752
 753     outeriter        = 0;
 754     inneriter        = 0;
 755
 756     /* Start outer loop over neighborlists */
 757     for(iidx=0; iidx<nri; iidx++)
 758     {
 759         /* Load shift vector for this list */
 760         i_shift_offset   = DIM*shiftidx[iidx];
 761
 762         /* Load limits for loop over neighbors */
 763         j_index_start    = jindex[iidx];
 764         j_index_end      = jindex[iidx+1];
 765
 766         /* Get outer coordinate index */
 767         inr              = iinr[iidx];
 768         i_coord_offset   = DIM*inr;
 769
 770         /* Load i particle coords and add shift vector */
 771         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 772                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 773
 774         fix0             = _fjsp_setzero_v2r8();
 775         fiy0             = _fjsp_setzero_v2r8();
 776         fiz0             = _fjsp_setzero_v2r8();
 777         fix1             = _fjsp_setzero_v2r8();
 778         fiy1             = _fjsp_setzero_v2r8();
 779         fiz1             = _fjsp_setzero_v2r8();
 780         fix2             = _fjsp_setzero_v2r8();
 781         fiy2             = _fjsp_setzero_v2r8();
 782         fiz2             = _fjsp_setzero_v2r8();
 783
 784         /* Start inner kernel loop */
 785         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 786         {
 787
 788             /* Get j neighbor index, and coordinate index */
 789             jnrA             = jjnr[jidx];
 790             jnrB             = jjnr[jidx+1];
 791             j_coord_offsetA  = DIM*jnrA;
 792             j_coord_offsetB  = DIM*jnrB;
 793
 794             /* load j atom coordinates */
 795             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 796                                               &jx0,&jy0,&jz0);
 797
 798             /* Calculate displacement vector */
 799             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 800             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 801             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 802             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 803             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 804             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 805             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 806             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 807             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 808
 809             /* Calculate squared distance and things based on it */
 810             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 811             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 812             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 813
 814             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 815             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 816             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 817
 818             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 819             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 820             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 821
 822             /* Load parameters for j particles */
 823             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 824             vdwjidx0A        = 2*vdwtype[jnrA+0];
 825             vdwjidx0B        = 2*vdwtype[jnrB+0];
 826
 827             fjx0             = _fjsp_setzero_v2r8();
 828             fjy0             = _fjsp_setzero_v2r8();
 829             fjz0             = _fjsp_setzero_v2r8();
 830
 831             /**************************
 832              * CALCULATE INTERACTIONS *
 833              **************************/
 834
 835             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 836             {
 837
 838             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 839
 840             /* Compute parameters for interactions between i and j atoms */
 841             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 842             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 843                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 844
 845             /* EWALD ELECTROSTATICS */
 846
 847             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 848             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 849             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 850             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 851             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 852
 853             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 854                                          &ewtabF,&ewtabFn);
 855             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 856             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 857
 858             /* LENNARD-JONES DISPERSION/REPULSION */
 859
 860             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 861             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 862
 863             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 864
 865             fscal            = _fjsp_add_v2r8(felec,fvdw);
 866
 867             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 868
 869             /* Update vectorial force */
 870             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 871             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 872             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 873
 874             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 875             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 876             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 877
 878             }
 879
 880             /**************************
 881              * CALCULATE INTERACTIONS *
 882              **************************/
 883
 884             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 885             {
 886
 887             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 888
 889             /* Compute parameters for interactions between i and j atoms */
 890             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 891
 892             /* EWALD ELECTROSTATICS */
 893
 894             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 895             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 896             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 897             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 898             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 899
 900             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 901                                          &ewtabF,&ewtabFn);
 902             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 903             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 904
 905             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 906
 907             fscal            = felec;
 908
 909             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 910
 911             /* Update vectorial force */
 912             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 913             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 914             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 915
 916             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 917             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 918             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 919
 920             }
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 927             {
 928
 929             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 930
 931             /* Compute parameters for interactions between i and j atoms */
 932             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 933
 934             /* EWALD ELECTROSTATICS */
 935
 936             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 937             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 938             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 939             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 940             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 941
 942             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 943                                          &ewtabF,&ewtabFn);
 944             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 945             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 946
 947             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 948
 949             fscal            = felec;
 950
 951             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 952
 953             /* Update vectorial force */
 954             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 955             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 956             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 957
 958             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 959             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 960             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 961
 962             }
 963
 964             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 965
 966             /* Inner loop uses 136 flops */
 967         }
 968
 969         if(jidx<j_index_end)
 970         {
 971
 972             jnrA             = jjnr[jidx];
 973             j_coord_offsetA  = DIM*jnrA;
 974
 975             /* load j atom coordinates */
 976             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 977                                               &jx0,&jy0,&jz0);
 978
 979             /* Calculate displacement vector */
 980             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 981             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 982             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 983             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 984             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 985             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 986             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 987             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 988             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 989
 990             /* Calculate squared distance and things based on it */
 991             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 992             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 993             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 994
 995             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 996             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 997             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 998
 999             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
1000             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1001             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1002
1003             /* Load parameters for j particles */
1004             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1005             vdwjidx0A        = 2*vdwtype[jnrA+0];
1006
1007             fjx0             = _fjsp_setzero_v2r8();
1008             fjy0             = _fjsp_setzero_v2r8();
1009             fjz0             = _fjsp_setzero_v2r8();
1010
1011             /**************************
1012              * CALCULATE INTERACTIONS *
1013              **************************/
1014
1015             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
1016             {
1017
1018             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1019
1020             /* Compute parameters for interactions between i and j atoms */
1021             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1022             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1023
1024             /* EWALD ELECTROSTATICS */
1025
1026             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1027             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1028             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1029             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1030             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1031
1032             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1033             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1034             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1035
1036             /* LENNARD-JONES DISPERSION/REPULSION */
1037
1038             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1039             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1040
1041             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
1042
1043             fscal            = _fjsp_add_v2r8(felec,fvdw);
1044
1045             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1046
1047             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1048
1049             /* Update vectorial force */
1050             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1051             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1052             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1053
1054             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1055             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1056             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1057
1058             }
1059
1060             /**************************
1061              * CALCULATE INTERACTIONS *
1062              **************************/
1063
1064             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1065             {
1066
1067             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1068
1069             /* Compute parameters for interactions between i and j atoms */
1070             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1071
1072             /* EWALD ELECTROSTATICS */
1073
1074             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1075             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1076             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1077             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1078             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1079
1080             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1081             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1082             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1083
1084             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1085
1086             fscal            = felec;
1087
1088             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1089
1090             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1091
1092             /* Update vectorial force */
1093             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1094             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1095             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1096
1097             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1098             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1099             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1100
1101             }
1102
1103             /**************************
1104              * CALCULATE INTERACTIONS *
1105              **************************/
1106
1107             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1108             {
1109
1110             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1111
1112             /* Compute parameters for interactions between i and j atoms */
1113             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1114
1115             /* EWALD ELECTROSTATICS */
1116
1117             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1118             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1119             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1120             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1121             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1122
1123             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1124             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1125             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1126
1127             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1128
1129             fscal            = felec;
1130
1131             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1132
1133             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1134
1135             /* Update vectorial force */
1136             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1137             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1138             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1139
1140             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1141             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1142             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1143
1144             }
1145
1146             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1147
1148             /* Inner loop uses 136 flops */
1149         }
1150
1151         /* End of innermost loop */
1152
1153         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1154                                               f+i_coord_offset,fshift+i_shift_offset);
1155
1156         /* Increment number of inner iterations */
1157         inneriter                  += j_index_end - j_index_start;
1158
1159         /* Outer loop uses 18 flops */
1160     }
1161
1162     /* Increment number of outer iterations */
1163     outeriter        += nri;
1164
1165     /* Update outer/inner flops */
1166
1167     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*136);
1168 }