src/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012, by the GROMACS development team, led by
   5  * David van der Spoel, Berk Hess, Erik Lindahl, and including many
   6  * others, as listed in the AUTHORS file in the top-level source
   7  * 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_ElecEw_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEw_VdwCSTab_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_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       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 101     real             *vftab;
 102     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 103     real             *ewtab;
 104     _fjsp_v2r8       itab_tmp;
 105     _fjsp_v2r8       dummy_mask,cutoff_mask;
 106     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 107     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 108     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 109
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123     nvdwtype         = fr->ntype;
 124     vdwparam         = fr->nbfp;
 125     vdwtype          = mdatoms->typeA;
 126
 127     vftab            = kernel_data->table_vdw->data;
 128     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 129
 130     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 131     ewtab            = fr->ic->tabq_coul_FDV0;
 132     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 133     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 134
 135     /* Setup water-specific parameters */
 136     inr              = nlist->iinr[0];
 137     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 138     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 139     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 140     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 141
 142     /* Avoid stupid compiler warnings */
 143     jnrA = jnrB = 0;
 144     j_coord_offsetA = 0;
 145     j_coord_offsetB = 0;
 146
 147     outeriter        = 0;
 148     inneriter        = 0;
 149
 150     /* Start outer loop over neighborlists */
 151     for(iidx=0; iidx<nri; iidx++)
 152     {
 153         /* Load shift vector for this list */
 154         i_shift_offset   = DIM*shiftidx[iidx];
 155
 156         /* Load limits for loop over neighbors */
 157         j_index_start    = jindex[iidx];
 158         j_index_end      = jindex[iidx+1];
 159
 160         /* Get outer coordinate index */
 161         inr              = iinr[iidx];
 162         i_coord_offset   = DIM*inr;
 163
 164         /* Load i particle coords and add shift vector */
 165         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 166                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 167
 168         fix0             = _fjsp_setzero_v2r8();
 169         fiy0             = _fjsp_setzero_v2r8();
 170         fiz0             = _fjsp_setzero_v2r8();
 171         fix1             = _fjsp_setzero_v2r8();
 172         fiy1             = _fjsp_setzero_v2r8();
 173         fiz1             = _fjsp_setzero_v2r8();
 174         fix2             = _fjsp_setzero_v2r8();
 175         fiy2             = _fjsp_setzero_v2r8();
 176         fiz2             = _fjsp_setzero_v2r8();
 177
 178         /* Reset potential sums */
 179         velecsum         = _fjsp_setzero_v2r8();
 180         vvdwsum          = _fjsp_setzero_v2r8();
 181
 182         /* Start inner kernel loop */
 183         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 184         {
 185
 186             /* Get j neighbor index, and coordinate index */
 187             jnrA             = jjnr[jidx];
 188             jnrB             = jjnr[jidx+1];
 189             j_coord_offsetA  = DIM*jnrA;
 190             j_coord_offsetB  = DIM*jnrB;
 191
 192             /* load j atom coordinates */
 193             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 194                                               &jx0,&jy0,&jz0);
 195
 196             /* Calculate displacement vector */
 197             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 198             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 199             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 200             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 201             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 202             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 203             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 204             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 205             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 206
 207             /* Calculate squared distance and things based on it */
 208             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 209             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 210             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 211
 212             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 213             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 214             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 215
 216             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 217             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 218             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 219
 220             /* Load parameters for j particles */
 221             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 222             vdwjidx0A        = 2*vdwtype[jnrA+0];
 223             vdwjidx0B        = 2*vdwtype[jnrB+0];
 224
 225             fjx0             = _fjsp_setzero_v2r8();
 226             fjy0             = _fjsp_setzero_v2r8();
 227             fjz0             = _fjsp_setzero_v2r8();
 228
 229             /**************************
 230              * CALCULATE INTERACTIONS *
 231              **************************/
 232
 233             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 234
 235             /* Compute parameters for interactions between i and j atoms */
 236             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 237             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 238                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 239
 240             /* Calculate table index by multiplying r with table scale and truncate to integer */
 241             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 242             itab_tmp         = _fjsp_dtox_v2r8(rt);
 243             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 244             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 245             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 246
 247             vfconv.i[0]     *= 8;
 248             vfconv.i[1]     *= 8;
 249
 250             /* EWALD ELECTROSTATICS */
 251
 252             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 253             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 254             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 255             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 256             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 257
 258             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 259             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 260             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 261             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 262             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 263             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 264             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 265             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 266             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 267             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 268
 269             /* CUBIC SPLINE TABLE DISPERSION */
 270             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 271             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 272             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 273             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 274             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 276             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 277             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 278             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 279             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 280             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 281
 282             /* CUBIC SPLINE TABLE REPULSION */
 283             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 284             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 285             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 286             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 287             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 288             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 289             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 290             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 291             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 292             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 293             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 294             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 295             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 296
 297             /* Update potential sum for this i atom from the interaction with this j atom. */
 298             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 299             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 300
 301             fscal            = _fjsp_add_v2r8(felec,fvdw);
 302
 303             /* Update vectorial force */
 304             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 305             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 306             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 307
 308             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 309             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 310             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 311
 312             /**************************
 313              * CALCULATE INTERACTIONS *
 314              **************************/
 315
 316             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 317
 318             /* Compute parameters for interactions between i and j atoms */
 319             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 320
 321             /* EWALD ELECTROSTATICS */
 322
 323             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 324             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 325             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 326             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 327             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 328
 329             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 330             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 331             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 332             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 333             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 334             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 335             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 336             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 337             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 338             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 339
 340             /* Update potential sum for this i atom from the interaction with this j atom. */
 341             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 342
 343             fscal            = felec;
 344
 345             /* Update vectorial force */
 346             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 347             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 348             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 349
 350             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 351             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 352             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 353
 354             /**************************
 355              * CALCULATE INTERACTIONS *
 356              **************************/
 357
 358             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 359
 360             /* Compute parameters for interactions between i and j atoms */
 361             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 362
 363             /* EWALD ELECTROSTATICS */
 364
 365             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 366             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 367             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 368             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 369             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 370
 371             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 372             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 373             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 374             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 375             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 376             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 377             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 378             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 379             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 380             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387             /* Update vectorial force */
 388             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 389             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 390             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 391
 392             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 393             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 394             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 395
 396             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 397
 398             /* Inner loop uses 169 flops */
 399         }
 400
 401         if(jidx<j_index_end)
 402         {
 403
 404             jnrA             = jjnr[jidx];
 405             j_coord_offsetA  = DIM*jnrA;
 406
 407             /* load j atom coordinates */
 408             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 409                                               &jx0,&jy0,&jz0);
 410
 411             /* Calculate displacement vector */
 412             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 413             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 414             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 415             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 416             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 417             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 418             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 419             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 420             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 421
 422             /* Calculate squared distance and things based on it */
 423             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 424             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 425             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 426
 427             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 428             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 429             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 430
 431             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 432             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 433             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 434
 435             /* Load parameters for j particles */
 436             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 437             vdwjidx0A        = 2*vdwtype[jnrA+0];
 438
 439             fjx0             = _fjsp_setzero_v2r8();
 440             fjy0             = _fjsp_setzero_v2r8();
 441             fjz0             = _fjsp_setzero_v2r8();
 442
 443             /**************************
 444              * CALCULATE INTERACTIONS *
 445              **************************/
 446
 447             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 448
 449             /* Compute parameters for interactions between i and j atoms */
 450             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 451             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 452
 453             /* Calculate table index by multiplying r with table scale and truncate to integer */
 454             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 455             itab_tmp         = _fjsp_dtox_v2r8(rt);
 456             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 457             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 458             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 459
 460             vfconv.i[0]     *= 8;
 461             vfconv.i[1]     *= 8;
 462
 463             /* EWALD ELECTROSTATICS */
 464
 465             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 466             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 467             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 468             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 469             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 470
 471             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 472             ewtabD           = _fjsp_setzero_v2r8();
 473             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 474             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 475             ewtabFn          = _fjsp_setzero_v2r8();
 476             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 477             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 478             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 479             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 480             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 481
 482             /* CUBIC SPLINE TABLE DISPERSION */
 483             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 484             F                = _fjsp_setzero_v2r8();
 485             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 486             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 487             H                = _fjsp_setzero_v2r8();
 488             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 489             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 490             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 491             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 492             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 493             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 494
 495             /* CUBIC SPLINE TABLE REPULSION */
 496             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 497             F                = _fjsp_setzero_v2r8();
 498             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 499             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 500             H                = _fjsp_setzero_v2r8();
 501             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 502             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 503             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 504             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 505             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 506             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 507             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 508             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 509
 510             /* Update potential sum for this i atom from the interaction with this j atom. */
 511             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 512             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 513             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 514             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 515
 516             fscal            = _fjsp_add_v2r8(felec,fvdw);
 517
 518             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 519
 520             /* Update vectorial force */
 521             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 522             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 523             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 524
 525             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 526             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 527             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 528
 529             /**************************
 530              * CALCULATE INTERACTIONS *
 531              **************************/
 532
 533             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 534
 535             /* Compute parameters for interactions between i and j atoms */
 536             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 537
 538             /* EWALD ELECTROSTATICS */
 539
 540             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 541             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 542             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 543             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 544             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 545
 546             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 547             ewtabD           = _fjsp_setzero_v2r8();
 548             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 549             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 550             ewtabFn          = _fjsp_setzero_v2r8();
 551             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 552             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 553             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 554             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 555             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 556
 557             /* Update potential sum for this i atom from the interaction with this j atom. */
 558             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 559             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 560
 561             fscal            = felec;
 562
 563             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 564
 565             /* Update vectorial force */
 566             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 567             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 568             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 569
 570             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 571             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 572             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 573
 574             /**************************
 575              * CALCULATE INTERACTIONS *
 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(rinv20,velec));
 600             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 601
 602             /* Update potential sum for this i atom from the interaction with this j atom. */
 603             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 604             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 605
 606             fscal            = felec;
 607
 608             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 609
 610             /* Update vectorial force */
 611             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 612             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 613             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 614
 615             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 616             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 617             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 618
 619             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 620
 621             /* Inner loop uses 169 flops */
 622         }
 623
 624         /* End of innermost loop */
 625
 626         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 627                                               f+i_coord_offset,fshift+i_shift_offset);
 628
 629         ggid                        = gid[iidx];
 630         /* Update potential energies */
 631         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 632         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 633
 634         /* Increment number of inner iterations */
 635         inneriter                  += j_index_end - j_index_start;
 636
 637         /* Outer loop uses 20 flops */
 638     }
 639
 640     /* Increment number of outer iterations */
 641     outeriter        += nri;
 642
 643     /* Update outer/inner flops */
 644
 645     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*169);
 646 }
 647 /*
 648  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 649  * Electrostatics interaction: Ewald
 650  * VdW interaction:            CubicSplineTable
 651  * Geometry:                   Water3-Particle
 652  * Calculate force/pot:        Force
 653  */
 654 void
 655 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 656                     (t_nblist * gmx_restrict                nlist,
 657                      rvec * gmx_restrict                    xx,
 658                      rvec * gmx_restrict                    ff,
 659                      t_forcerec * gmx_restrict              fr,
 660                      t_mdatoms * gmx_restrict               mdatoms,
 661                      nb_kernel_data_t * gmx_restrict        kernel_data,
 662                      t_nrnb * gmx_restrict                  nrnb)
 663 {
 664     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 665      * just 0 for non-waters.
 666      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 667      * jnr indices corresponding to data put in the four positions in the SIMD register.
 668      */
 669     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 670     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 671     int              jnrA,jnrB;
 672     int              j_coord_offsetA,j_coord_offsetB;
 673     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 674     real             rcutoff_scalar;
 675     real             *shiftvec,*fshift,*x,*f;
 676     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 677     int              vdwioffset0;
 678     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 679     int              vdwioffset1;
 680     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 681     int              vdwioffset2;
 682     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 683     int              vdwjidx0A,vdwjidx0B;
 684     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 685     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 686     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 687     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 688     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 689     real             *charge;
 690     int              nvdwtype;
 691     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 692     int              *vdwtype;
 693     real             *vdwparam;
 694     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 695     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 696     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 697     real             *vftab;
 698     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 699     real             *ewtab;
 700     _fjsp_v2r8       itab_tmp;
 701     _fjsp_v2r8       dummy_mask,cutoff_mask;
 702     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 703     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 704     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 705
 706     x                = xx[0];
 707     f                = ff[0];
 708
 709     nri              = nlist->nri;
 710     iinr             = nlist->iinr;
 711     jindex           = nlist->jindex;
 712     jjnr             = nlist->jjnr;
 713     shiftidx         = nlist->shift;
 714     gid              = nlist->gid;
 715     shiftvec         = fr->shift_vec[0];
 716     fshift           = fr->fshift[0];
 717     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 718     charge           = mdatoms->chargeA;
 719     nvdwtype         = fr->ntype;
 720     vdwparam         = fr->nbfp;
 721     vdwtype          = mdatoms->typeA;
 722
 723     vftab            = kernel_data->table_vdw->data;
 724     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 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     /* Avoid stupid compiler warnings */
 739     jnrA = jnrB = 0;
 740     j_coord_offsetA = 0;
 741     j_coord_offsetB = 0;
 742
 743     outeriter        = 0;
 744     inneriter        = 0;
 745
 746     /* Start outer loop over neighborlists */
 747     for(iidx=0; iidx<nri; iidx++)
 748     {
 749         /* Load shift vector for this list */
 750         i_shift_offset   = DIM*shiftidx[iidx];
 751
 752         /* Load limits for loop over neighbors */
 753         j_index_start    = jindex[iidx];
 754         j_index_end      = jindex[iidx+1];
 755
 756         /* Get outer coordinate index */
 757         inr              = iinr[iidx];
 758         i_coord_offset   = DIM*inr;
 759
 760         /* Load i particle coords and add shift vector */
 761         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 762                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 763
 764         fix0             = _fjsp_setzero_v2r8();
 765         fiy0             = _fjsp_setzero_v2r8();
 766         fiz0             = _fjsp_setzero_v2r8();
 767         fix1             = _fjsp_setzero_v2r8();
 768         fiy1             = _fjsp_setzero_v2r8();
 769         fiz1             = _fjsp_setzero_v2r8();
 770         fix2             = _fjsp_setzero_v2r8();
 771         fiy2             = _fjsp_setzero_v2r8();
 772         fiz2             = _fjsp_setzero_v2r8();
 773
 774         /* Start inner kernel loop */
 775         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 776         {
 777
 778             /* Get j neighbor index, and coordinate index */
 779             jnrA             = jjnr[jidx];
 780             jnrB             = jjnr[jidx+1];
 781             j_coord_offsetA  = DIM*jnrA;
 782             j_coord_offsetB  = DIM*jnrB;
 783
 784             /* load j atom coordinates */
 785             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 786                                               &jx0,&jy0,&jz0);
 787
 788             /* Calculate displacement vector */
 789             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 790             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 791             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 792             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 793             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 794             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 795             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 796             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 797             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 798
 799             /* Calculate squared distance and things based on it */
 800             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 801             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 802             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 803
 804             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 805             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 806             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 807
 808             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 809             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 810             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 811
 812             /* Load parameters for j particles */
 813             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 814             vdwjidx0A        = 2*vdwtype[jnrA+0];
 815             vdwjidx0B        = 2*vdwtype[jnrB+0];
 816
 817             fjx0             = _fjsp_setzero_v2r8();
 818             fjy0             = _fjsp_setzero_v2r8();
 819             fjz0             = _fjsp_setzero_v2r8();
 820
 821             /**************************
 822              * CALCULATE INTERACTIONS *
 823              **************************/
 824
 825             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 826
 827             /* Compute parameters for interactions between i and j atoms */
 828             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 829             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 830                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 831
 832             /* Calculate table index by multiplying r with table scale and truncate to integer */
 833             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 834             itab_tmp         = _fjsp_dtox_v2r8(rt);
 835             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 836             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 837             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 838
 839             vfconv.i[0]     *= 8;
 840             vfconv.i[1]     *= 8;
 841
 842             /* EWALD ELECTROSTATICS */
 843
 844             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 845             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 846             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 847             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 848             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 849
 850             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 851                                          &ewtabF,&ewtabFn);
 852             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 853             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 854
 855             /* CUBIC SPLINE TABLE DISPERSION */
 856             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 857             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 858             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 859             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 860             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 861             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 862             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 863             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 864             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 865
 866             /* CUBIC SPLINE TABLE REPULSION */
 867             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 868             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 869             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 870             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 871             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 872             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 873             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 874             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 875             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 876             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 877
 878             fscal            = _fjsp_add_v2r8(felec,fvdw);
 879
 880             /* Update vectorial force */
 881             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 882             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 883             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 884
 885             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 886             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 887             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 888
 889             /**************************
 890              * CALCULATE INTERACTIONS *
 891              **************************/
 892
 893             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 894
 895             /* Compute parameters for interactions between i and j atoms */
 896             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 897
 898             /* EWALD ELECTROSTATICS */
 899
 900             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 901             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 902             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 903             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 904             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 905
 906             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 907                                          &ewtabF,&ewtabFn);
 908             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 909             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 910
 911             fscal            = felec;
 912
 913             /* Update vectorial force */
 914             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 915             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 916             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 917
 918             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 919             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 920             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 921
 922             /**************************
 923              * CALCULATE INTERACTIONS *
 924              **************************/
 925
 926             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 927
 928             /* Compute parameters for interactions between i and j atoms */
 929             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 930
 931             /* EWALD ELECTROSTATICS */
 932
 933             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 934             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 935             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 936             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 937             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 938
 939             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 940                                          &ewtabF,&ewtabFn);
 941             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 942             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 943
 944             fscal            = felec;
 945
 946             /* Update vectorial force */
 947             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 948             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 949             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 950
 951             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 952             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 953             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 954
 955             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 956
 957             /* Inner loop uses 146 flops */
 958         }
 959
 960         if(jidx<j_index_end)
 961         {
 962
 963             jnrA             = jjnr[jidx];
 964             j_coord_offsetA  = DIM*jnrA;
 965
 966             /* load j atom coordinates */
 967             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 968                                               &jx0,&jy0,&jz0);
 969
 970             /* Calculate displacement vector */
 971             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 972             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 973             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 974             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 975             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 976             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 977             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 978             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 979             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 980
 981             /* Calculate squared distance and things based on it */
 982             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 983             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 984             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 985
 986             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 987             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 988             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 989
 990             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 991             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 992             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 993
 994             /* Load parameters for j particles */
 995             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 996             vdwjidx0A        = 2*vdwtype[jnrA+0];
 997
 998             fjx0             = _fjsp_setzero_v2r8();
 999             fjy0             = _fjsp_setzero_v2r8();
1000             fjz0             = _fjsp_setzero_v2r8();
1001
1002             /**************************
1003              * CALCULATE INTERACTIONS *
1004              **************************/
1005
1006             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1007
1008             /* Compute parameters for interactions between i and j atoms */
1009             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1010             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1011
1012             /* Calculate table index by multiplying r with table scale and truncate to integer */
1013             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1014             itab_tmp         = _fjsp_dtox_v2r8(rt);
1015             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1016             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1017             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1018
1019             vfconv.i[0]     *= 8;
1020             vfconv.i[1]     *= 8;
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             /* CUBIC SPLINE TABLE DISPERSION */
1035             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1036             F                = _fjsp_setzero_v2r8();
1037             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1038             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1039             H                = _fjsp_setzero_v2r8();
1040             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1041             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1042             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1043             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1044
1045             /* CUBIC SPLINE TABLE REPULSION */
1046             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1047             F                = _fjsp_setzero_v2r8();
1048             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1049             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1050             H                = _fjsp_setzero_v2r8();
1051             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1052             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1053             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1054             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1055             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1056
1057             fscal            = _fjsp_add_v2r8(felec,fvdw);
1058
1059             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1060
1061             /* Update vectorial force */
1062             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1063             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1064             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1065
1066             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1067             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1068             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1069
1070             /**************************
1071              * CALCULATE INTERACTIONS *
1072              **************************/
1073
1074             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1075
1076             /* Compute parameters for interactions between i and j atoms */
1077             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1078
1079             /* EWALD ELECTROSTATICS */
1080
1081             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1082             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1083             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1084             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1085             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1086
1087             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1088             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1089             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1090
1091             fscal            = felec;
1092
1093             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1094
1095             /* Update vectorial force */
1096             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1097             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1098             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1099
1100             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1101             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1102             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1103
1104             /**************************
1105              * CALCULATE INTERACTIONS *
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             fscal            = felec;
1126
1127             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1128
1129             /* Update vectorial force */
1130             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1131             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1132             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1133
1134             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1135             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1136             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1137
1138             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1139
1140             /* Inner loop uses 146 flops */
1141         }
1142
1143         /* End of innermost loop */
1144
1145         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1146                                               f+i_coord_offset,fshift+i_shift_offset);
1147
1148         /* Increment number of inner iterations */
1149         inneriter                  += j_index_end - j_index_start;
1150
1151         /* Outer loop uses 18 flops */
1152     }
1153
1154     /* Increment number of outer iterations */
1155     outeriter        += nri;
1156
1157     /* Update outer/inner flops */
1158
1159     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*146);
1160 }