src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2017,2018, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/gmxlib/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      struct t_forcerec           * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 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->ic->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_vdw->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 127
 128     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 129     ewtab            = fr->ic->tabq_coul_FDV0;
 130     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 131     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 132
 133     /* Setup water-specific parameters */
 134     inr              = nlist->iinr[0];
 135     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 136     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 137     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 138     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 139
 140     /* Avoid stupid compiler warnings */
 141     jnrA = jnrB = 0;
 142     j_coord_offsetA = 0;
 143     j_coord_offsetB = 0;
 144
 145     outeriter        = 0;
 146     inneriter        = 0;
 147
 148     /* Start outer loop over neighborlists */
 149     for(iidx=0; iidx<nri; iidx++)
 150     {
 151         /* Load shift vector for this list */
 152         i_shift_offset   = DIM*shiftidx[iidx];
 153
 154         /* Load limits for loop over neighbors */
 155         j_index_start    = jindex[iidx];
 156         j_index_end      = jindex[iidx+1];
 157
 158         /* Get outer coordinate index */
 159         inr              = iinr[iidx];
 160         i_coord_offset   = DIM*inr;
 161
 162         /* Load i particle coords and add shift vector */
 163         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 164                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 165
 166         fix0             = _fjsp_setzero_v2r8();
 167         fiy0             = _fjsp_setzero_v2r8();
 168         fiz0             = _fjsp_setzero_v2r8();
 169         fix1             = _fjsp_setzero_v2r8();
 170         fiy1             = _fjsp_setzero_v2r8();
 171         fiz1             = _fjsp_setzero_v2r8();
 172         fix2             = _fjsp_setzero_v2r8();
 173         fiy2             = _fjsp_setzero_v2r8();
 174         fiz2             = _fjsp_setzero_v2r8();
 175
 176         /* Reset potential sums */
 177         velecsum         = _fjsp_setzero_v2r8();
 178         vvdwsum          = _fjsp_setzero_v2r8();
 179
 180         /* Start inner kernel loop */
 181         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 182         {
 183
 184             /* Get j neighbor index, and coordinate index */
 185             jnrA             = jjnr[jidx];
 186             jnrB             = jjnr[jidx+1];
 187             j_coord_offsetA  = DIM*jnrA;
 188             j_coord_offsetB  = DIM*jnrB;
 189
 190             /* load j atom coordinates */
 191             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 192                                               &jx0,&jy0,&jz0);
 193
 194             /* Calculate displacement vector */
 195             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 196             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 197             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 198             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 199             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 200             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 201             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 202             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 203             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 204
 205             /* Calculate squared distance and things based on it */
 206             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 207             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 208             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 209
 210             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 211             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 212             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 213
 214             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 215             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 216             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 217
 218             /* Load parameters for j particles */
 219             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 220             vdwjidx0A        = 2*vdwtype[jnrA+0];
 221             vdwjidx0B        = 2*vdwtype[jnrB+0];
 222
 223             fjx0             = _fjsp_setzero_v2r8();
 224             fjy0             = _fjsp_setzero_v2r8();
 225             fjz0             = _fjsp_setzero_v2r8();
 226
 227             /**************************
 228              * CALCULATE INTERACTIONS *
 229              **************************/
 230
 231             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 232
 233             /* Compute parameters for interactions between i and j atoms */
 234             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 235             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 236                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 237
 238             /* Calculate table index by multiplying r with table scale and truncate to integer */
 239             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 240             itab_tmp         = _fjsp_dtox_v2r8(rt);
 241             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 242             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 243             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 244
 245             vfconv.i[0]     *= 8;
 246             vfconv.i[1]     *= 8;
 247
 248             /* EWALD ELECTROSTATICS */
 249
 250             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 251             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 252             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 253             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 254             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 255
 256             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 257             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 258             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 259             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 260             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 261             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 262             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 263             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 264             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 265             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 266
 267             /* CUBIC SPLINE TABLE DISPERSION */
 268             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 269             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 270             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 271             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 272             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 273             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 274             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 275             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 276             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 277             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 278             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 279
 280             /* CUBIC SPLINE TABLE REPULSION */
 281             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 282             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 283             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 284             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 285             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 286             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 287             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 288             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 289             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 290             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 291             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 292             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 293             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 294
 295             /* Update potential sum for this i atom from the interaction with this j atom. */
 296             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 297             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 298
 299             fscal            = _fjsp_add_v2r8(felec,fvdw);
 300
 301             /* Update vectorial force */
 302             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 303             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 304             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 305
 306             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 307             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 308             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 309
 310             /**************************
 311              * CALCULATE INTERACTIONS *
 312              **************************/
 313
 314             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 315
 316             /* Compute parameters for interactions between i and j atoms */
 317             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 318
 319             /* EWALD ELECTROSTATICS */
 320
 321             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 322             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 323             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 324             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 325             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 326
 327             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 328             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 329             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 330             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 331             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 332             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 333             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 334             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 335             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 336             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 340
 341             fscal            = felec;
 342
 343             /* Update vectorial force */
 344             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 345             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 346             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 347
 348             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 349             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 350             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 351
 352             /**************************
 353              * CALCULATE INTERACTIONS *
 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(rinv20,velec));
 378             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 379
 380             /* Update potential sum for this i atom from the interaction with this j atom. */
 381             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 382
 383             fscal            = felec;
 384
 385             /* Update vectorial force */
 386             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 387             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 388             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 389
 390             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 391             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 392             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 393
 394             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 395
 396             /* Inner loop uses 169 flops */
 397         }
 398
 399         if(jidx<j_index_end)
 400         {
 401
 402             jnrA             = jjnr[jidx];
 403             j_coord_offsetA  = DIM*jnrA;
 404
 405             /* load j atom coordinates */
 406             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 407                                               &jx0,&jy0,&jz0);
 408
 409             /* Calculate displacement vector */
 410             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 411             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 412             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 413             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 414             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 415             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 416             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 417             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 418             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 419
 420             /* Calculate squared distance and things based on it */
 421             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 422             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 423             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 424
 425             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 426             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 427             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 428
 429             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 430             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 431             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 432
 433             /* Load parameters for j particles */
 434             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 435             vdwjidx0A        = 2*vdwtype[jnrA+0];
 436
 437             fjx0             = _fjsp_setzero_v2r8();
 438             fjy0             = _fjsp_setzero_v2r8();
 439             fjz0             = _fjsp_setzero_v2r8();
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 446
 447             /* Compute parameters for interactions between i and j atoms */
 448             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 449             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 450                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 451
 452             /* Calculate table index by multiplying r with table scale and truncate to integer */
 453             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 454             itab_tmp         = _fjsp_dtox_v2r8(rt);
 455             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 456             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 457             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 458
 459             vfconv.i[0]     *= 8;
 460             vfconv.i[1]     *= 8;
 461
 462             /* EWALD ELECTROSTATICS */
 463
 464             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 465             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 466             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 467             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 468             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 469
 470             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 471             ewtabD           = _fjsp_setzero_v2r8();
 472             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 473             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 474             ewtabFn          = _fjsp_setzero_v2r8();
 475             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 476             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 477             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 478             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 479             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 480
 481             /* CUBIC SPLINE TABLE DISPERSION */
 482             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 483             F                = _fjsp_setzero_v2r8();
 484             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 485             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 486             H                = _fjsp_setzero_v2r8();
 487             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 488             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 489             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 490             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 491             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 492             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 493
 494             /* CUBIC SPLINE TABLE REPULSION */
 495             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 496             F                = _fjsp_setzero_v2r8();
 497             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 498             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 499             H                = _fjsp_setzero_v2r8();
 500             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 501             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 502             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 503             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 504             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 505             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 506             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 507             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 508
 509             /* Update potential sum for this i atom from the interaction with this j atom. */
 510             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 511             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 512             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 513             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 514
 515             fscal            = _fjsp_add_v2r8(felec,fvdw);
 516
 517             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 518
 519             /* Update vectorial force */
 520             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 521             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 522             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 523
 524             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 525             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 526             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 527
 528             /**************************
 529              * CALCULATE INTERACTIONS *
 530              **************************/
 531
 532             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 533
 534             /* Compute parameters for interactions between i and j atoms */
 535             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 536
 537             /* EWALD ELECTROSTATICS */
 538
 539             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 540             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 541             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 542             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 543             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 544
 545             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 546             ewtabD           = _fjsp_setzero_v2r8();
 547             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 548             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 549             ewtabFn          = _fjsp_setzero_v2r8();
 550             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 551             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 552             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 553             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 554             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 555
 556             /* Update potential sum for this i atom from the interaction with this j atom. */
 557             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 558             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 559
 560             fscal            = felec;
 561
 562             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 563
 564             /* Update vectorial force */
 565             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 566             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 567             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 568
 569             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 570             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 571             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 572
 573             /**************************
 574              * CALCULATE INTERACTIONS *
 575              **************************/
 576
 577             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 578
 579             /* Compute parameters for interactions between i and j atoms */
 580             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 581
 582             /* EWALD ELECTROSTATICS */
 583
 584             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 585             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 586             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 587             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 588             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 589
 590             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 591             ewtabD           = _fjsp_setzero_v2r8();
 592             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 593             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 594             ewtabFn          = _fjsp_setzero_v2r8();
 595             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 596             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 597             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 598             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 599             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 600
 601             /* Update potential sum for this i atom from the interaction with this j atom. */
 602             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 603             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 604
 605             fscal            = felec;
 606
 607             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 608
 609             /* Update vectorial force */
 610             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 611             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 612             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 613
 614             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 615             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 616             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 617
 618             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 619
 620             /* Inner loop uses 169 flops */
 621         }
 622
 623         /* End of innermost loop */
 624
 625         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 626                                               f+i_coord_offset,fshift+i_shift_offset);
 627
 628         ggid                        = gid[iidx];
 629         /* Update potential energies */
 630         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 631         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 632
 633         /* Increment number of inner iterations */
 634         inneriter                  += j_index_end - j_index_start;
 635
 636         /* Outer loop uses 20 flops */
 637     }
 638
 639     /* Increment number of outer iterations */
 640     outeriter        += nri;
 641
 642     /* Update outer/inner flops */
 643
 644     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*169);
 645 }
 646 /*
 647  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 648  * Electrostatics interaction: Ewald
 649  * VdW interaction:            CubicSplineTable
 650  * Geometry:                   Water3-Particle
 651  * Calculate force/pot:        Force
 652  */
 653 void
 654 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 655                     (t_nblist                    * gmx_restrict       nlist,
 656                      rvec                        * gmx_restrict          xx,
 657                      rvec                        * gmx_restrict          ff,
 658                      struct t_forcerec           * gmx_restrict          fr,
 659                      t_mdatoms                   * gmx_restrict     mdatoms,
 660                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 661                      t_nrnb                      * gmx_restrict        nrnb)
 662 {
 663     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 664      * just 0 for non-waters.
 665      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 666      * jnr indices corresponding to data put in the four positions in the SIMD register.
 667      */
 668     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 669     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 670     int              jnrA,jnrB;
 671     int              j_coord_offsetA,j_coord_offsetB;
 672     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 673     real             rcutoff_scalar;
 674     real             *shiftvec,*fshift,*x,*f;
 675     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 676     int              vdwioffset0;
 677     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 678     int              vdwioffset1;
 679     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 680     int              vdwioffset2;
 681     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 682     int              vdwjidx0A,vdwjidx0B;
 683     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 684     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 685     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 686     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 687     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 688     real             *charge;
 689     int              nvdwtype;
 690     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 691     int              *vdwtype;
 692     real             *vdwparam;
 693     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 694     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 695     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 696     real             *vftab;
 697     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 698     real             *ewtab;
 699     _fjsp_v2r8       itab_tmp;
 700     _fjsp_v2r8       dummy_mask,cutoff_mask;
 701     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 702     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 703     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 704
 705     x                = xx[0];
 706     f                = ff[0];
 707
 708     nri              = nlist->nri;
 709     iinr             = nlist->iinr;
 710     jindex           = nlist->jindex;
 711     jjnr             = nlist->jjnr;
 712     shiftidx         = nlist->shift;
 713     gid              = nlist->gid;
 714     shiftvec         = fr->shift_vec[0];
 715     fshift           = fr->fshift[0];
 716     facel            = gmx_fjsp_set1_v2r8(fr->ic->epsfac);
 717     charge           = mdatoms->chargeA;
 718     nvdwtype         = fr->ntype;
 719     vdwparam         = fr->nbfp;
 720     vdwtype          = mdatoms->typeA;
 721
 722     vftab            = kernel_data->table_vdw->data;
 723     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 724
 725     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 726     ewtab            = fr->ic->tabq_coul_F;
 727     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 728     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 729
 730     /* Setup water-specific parameters */
 731     inr              = nlist->iinr[0];
 732     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 733     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 734     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 735     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 736
 737     /* Avoid stupid compiler warnings */
 738     jnrA = jnrB = 0;
 739     j_coord_offsetA = 0;
 740     j_coord_offsetB = 0;
 741
 742     outeriter        = 0;
 743     inneriter        = 0;
 744
 745     /* Start outer loop over neighborlists */
 746     for(iidx=0; iidx<nri; iidx++)
 747     {
 748         /* Load shift vector for this list */
 749         i_shift_offset   = DIM*shiftidx[iidx];
 750
 751         /* Load limits for loop over neighbors */
 752         j_index_start    = jindex[iidx];
 753         j_index_end      = jindex[iidx+1];
 754
 755         /* Get outer coordinate index */
 756         inr              = iinr[iidx];
 757         i_coord_offset   = DIM*inr;
 758
 759         /* Load i particle coords and add shift vector */
 760         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 761                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 762
 763         fix0             = _fjsp_setzero_v2r8();
 764         fiy0             = _fjsp_setzero_v2r8();
 765         fiz0             = _fjsp_setzero_v2r8();
 766         fix1             = _fjsp_setzero_v2r8();
 767         fiy1             = _fjsp_setzero_v2r8();
 768         fiz1             = _fjsp_setzero_v2r8();
 769         fix2             = _fjsp_setzero_v2r8();
 770         fiy2             = _fjsp_setzero_v2r8();
 771         fiz2             = _fjsp_setzero_v2r8();
 772
 773         /* Start inner kernel loop */
 774         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 775         {
 776
 777             /* Get j neighbor index, and coordinate index */
 778             jnrA             = jjnr[jidx];
 779             jnrB             = jjnr[jidx+1];
 780             j_coord_offsetA  = DIM*jnrA;
 781             j_coord_offsetB  = DIM*jnrB;
 782
 783             /* load j atom coordinates */
 784             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 785                                               &jx0,&jy0,&jz0);
 786
 787             /* Calculate displacement vector */
 788             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 789             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 790             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 791             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 792             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 793             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 794             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 795             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 796             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 797
 798             /* Calculate squared distance and things based on it */
 799             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 800             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 801             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 802
 803             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 804             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 805             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 806
 807             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 808             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 809             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 810
 811             /* Load parameters for j particles */
 812             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 813             vdwjidx0A        = 2*vdwtype[jnrA+0];
 814             vdwjidx0B        = 2*vdwtype[jnrB+0];
 815
 816             fjx0             = _fjsp_setzero_v2r8();
 817             fjy0             = _fjsp_setzero_v2r8();
 818             fjz0             = _fjsp_setzero_v2r8();
 819
 820             /**************************
 821              * CALCULATE INTERACTIONS *
 822              **************************/
 823
 824             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 825
 826             /* Compute parameters for interactions between i and j atoms */
 827             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 828             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 829                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 830
 831             /* Calculate table index by multiplying r with table scale and truncate to integer */
 832             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 833             itab_tmp         = _fjsp_dtox_v2r8(rt);
 834             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 835             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 836             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 837
 838             vfconv.i[0]     *= 8;
 839             vfconv.i[1]     *= 8;
 840
 841             /* EWALD ELECTROSTATICS */
 842
 843             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 844             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 845             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 846             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 847             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 848
 849             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 850                                          &ewtabF,&ewtabFn);
 851             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 852             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 853
 854             /* CUBIC SPLINE TABLE DISPERSION */
 855             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 856             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 857             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 858             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 859             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 860             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 861             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 862             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 863             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 864
 865             /* CUBIC SPLINE TABLE REPULSION */
 866             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 867             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 868             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 869             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 870             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 871             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 872             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 873             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 874             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 875             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 876
 877             fscal            = _fjsp_add_v2r8(felec,fvdw);
 878
 879             /* Update vectorial force */
 880             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 881             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 882             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 883
 884             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 885             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 886             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 887
 888             /**************************
 889              * CALCULATE INTERACTIONS *
 890              **************************/
 891
 892             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 893
 894             /* Compute parameters for interactions between i and j atoms */
 895             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 896
 897             /* EWALD ELECTROSTATICS */
 898
 899             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 900             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 901             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 902             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 903             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 904
 905             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 906                                          &ewtabF,&ewtabFn);
 907             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 908             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 909
 910             fscal            = felec;
 911
 912             /* Update vectorial force */
 913             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 914             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 915             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 916
 917             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 918             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 919             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 920
 921             /**************************
 922              * CALCULATE INTERACTIONS *
 923              **************************/
 924
 925             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 926
 927             /* Compute parameters for interactions between i and j atoms */
 928             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 929
 930             /* EWALD ELECTROSTATICS */
 931
 932             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 933             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 934             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 935             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 936             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 937
 938             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 939                                          &ewtabF,&ewtabFn);
 940             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 941             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 942
 943             fscal            = felec;
 944
 945             /* Update vectorial force */
 946             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 947             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 948             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 949
 950             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 951             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 952             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 953
 954             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 955
 956             /* Inner loop uses 146 flops */
 957         }
 958
 959         if(jidx<j_index_end)
 960         {
 961
 962             jnrA             = jjnr[jidx];
 963             j_coord_offsetA  = DIM*jnrA;
 964
 965             /* load j atom coordinates */
 966             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 967                                               &jx0,&jy0,&jz0);
 968
 969             /* Calculate displacement vector */
 970             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 971             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 972             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 973             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 974             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 975             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 976             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 977             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 978             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 979
 980             /* Calculate squared distance and things based on it */
 981             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 982             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 983             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 984
 985             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 986             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 987             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 988
 989             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 990             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 991             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 992
 993             /* Load parameters for j particles */
 994             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 995             vdwjidx0A        = 2*vdwtype[jnrA+0];
 996
 997             fjx0             = _fjsp_setzero_v2r8();
 998             fjy0             = _fjsp_setzero_v2r8();
 999             fjz0             = _fjsp_setzero_v2r8();
1000
1001             /**************************
1002              * CALCULATE INTERACTIONS *
1003              **************************/
1004
1005             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1006
1007             /* Compute parameters for interactions between i and j atoms */
1008             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1009             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1010                                          vdwparam+vdwioffset0+vdwjidx0B,&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 }