src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_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_ElecCSTab_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_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_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 101     real             *vftab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_elec_vdw->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     /* Avoid stupid compiler warnings */
 136     jnrA = jnrB = 0;
 137     j_coord_offsetA = 0;
 138     j_coord_offsetB = 0;
 139
 140     outeriter        = 0;
 141     inneriter        = 0;
 142
 143     /* Start outer loop over neighborlists */
 144     for(iidx=0; iidx<nri; iidx++)
 145     {
 146         /* Load shift vector for this list */
 147         i_shift_offset   = DIM*shiftidx[iidx];
 148
 149         /* Load limits for loop over neighbors */
 150         j_index_start    = jindex[iidx];
 151         j_index_end      = jindex[iidx+1];
 152
 153         /* Get outer coordinate index */
 154         inr              = iinr[iidx];
 155         i_coord_offset   = DIM*inr;
 156
 157         /* Load i particle coords and add shift vector */
 158         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 159                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 160
 161         fix0             = _fjsp_setzero_v2r8();
 162         fiy0             = _fjsp_setzero_v2r8();
 163         fiz0             = _fjsp_setzero_v2r8();
 164         fix1             = _fjsp_setzero_v2r8();
 165         fiy1             = _fjsp_setzero_v2r8();
 166         fiz1             = _fjsp_setzero_v2r8();
 167         fix2             = _fjsp_setzero_v2r8();
 168         fiy2             = _fjsp_setzero_v2r8();
 169         fiz2             = _fjsp_setzero_v2r8();
 170
 171         /* Reset potential sums */
 172         velecsum         = _fjsp_setzero_v2r8();
 173         vvdwsum          = _fjsp_setzero_v2r8();
 174
 175         /* Start inner kernel loop */
 176         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 177         {
 178
 179             /* Get j neighbor index, and coordinate index */
 180             jnrA             = jjnr[jidx];
 181             jnrB             = jjnr[jidx+1];
 182             j_coord_offsetA  = DIM*jnrA;
 183             j_coord_offsetB  = DIM*jnrB;
 184
 185             /* load j atom coordinates */
 186             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 187                                               &jx0,&jy0,&jz0);
 188
 189             /* Calculate displacement vector */
 190             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 191             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 192             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 193             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 194             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 195             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 196             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 197             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 198             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 199
 200             /* Calculate squared distance and things based on it */
 201             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 202             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 203             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 204
 205             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 206             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 207             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 208
 209             /* Load parameters for j particles */
 210             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 211             vdwjidx0A        = 2*vdwtype[jnrA+0];
 212             vdwjidx0B        = 2*vdwtype[jnrB+0];
 213
 214             fjx0             = _fjsp_setzero_v2r8();
 215             fjy0             = _fjsp_setzero_v2r8();
 216             fjz0             = _fjsp_setzero_v2r8();
 217
 218             /**************************
 219              * CALCULATE INTERACTIONS *
 220              **************************/
 221
 222             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 223
 224             /* Compute parameters for interactions between i and j atoms */
 225             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 226             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 227                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 228
 229             /* Calculate table index by multiplying r with table scale and truncate to integer */
 230             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 231             itab_tmp         = _fjsp_dtox_v2r8(rt);
 232             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 233             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 234             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 235
 236             vfconv.i[0]     *= 12;
 237             vfconv.i[1]     *= 12;
 238
 239             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 240             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 241             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 242             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 243             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 244             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 245             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 246             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 247             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 248             velec            = _fjsp_mul_v2r8(qq00,VV);
 249             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 250             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 251
 252             /* CUBIC SPLINE TABLE DISPERSION */
 253             vfconv.i[0]       += 4;
 254             vfconv.i[1]       += 4;
 255             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 256             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 257             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 258             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 259             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 260             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 261             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 262             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 263             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 264             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 265             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 266
 267             /* CUBIC SPLINE TABLE REPULSION */
 268             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 269             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 271             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 272             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 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             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 277             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 278             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 279             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 280             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 281
 282             /* Update potential sum for this i atom from the interaction with this j atom. */
 283             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 284             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 285
 286             fscal            = _fjsp_add_v2r8(felec,fvdw);
 287
 288             /* Update vectorial force */
 289             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 290             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 291             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 292
 293             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 294             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 295             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 302
 303             /* Compute parameters for interactions between i and j atoms */
 304             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 305
 306             /* Calculate table index by multiplying r with table scale and truncate to integer */
 307             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 308             itab_tmp         = _fjsp_dtox_v2r8(rt);
 309             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 310             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 311             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 312
 313             vfconv.i[0]     *= 12;
 314             vfconv.i[1]     *= 12;
 315
 316             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 317             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 318             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 319             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 320             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 321             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 322             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 323             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 324             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 325             velec            = _fjsp_mul_v2r8(qq10,VV);
 326             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 327             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Update vectorial force */
 335             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 336             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 337             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 338
 339             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 340             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 341             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 342
 343             /**************************
 344              * CALCULATE INTERACTIONS *
 345              **************************/
 346
 347             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 348
 349             /* Compute parameters for interactions between i and j atoms */
 350             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 351
 352             /* Calculate table index by multiplying r with table scale and truncate to integer */
 353             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 354             itab_tmp         = _fjsp_dtox_v2r8(rt);
 355             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 356             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 357             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 358
 359             vfconv.i[0]     *= 12;
 360             vfconv.i[1]     *= 12;
 361
 362             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 363             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 364             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 365             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 366             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 367             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 368             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 369             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 370             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 371             velec            = _fjsp_mul_v2r8(qq20,VV);
 372             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 373             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 374
 375             /* Update potential sum for this i atom from the interaction with this j atom. */
 376             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 377
 378             fscal            = felec;
 379
 380             /* Update vectorial force */
 381             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 382             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 383             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 384
 385             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 386             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 387             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 388
 389             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 390
 391             /* Inner loop uses 171 flops */
 392         }
 393
 394         if(jidx<j_index_end)
 395         {
 396
 397             jnrA             = jjnr[jidx];
 398             j_coord_offsetA  = DIM*jnrA;
 399
 400             /* load j atom coordinates */
 401             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 402                                               &jx0,&jy0,&jz0);
 403
 404             /* Calculate displacement vector */
 405             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 406             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 407             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 408             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 409             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 410             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 411             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 412             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 413             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 414
 415             /* Calculate squared distance and things based on it */
 416             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 417             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 418             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 419
 420             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 421             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 422             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 423
 424             /* Load parameters for j particles */
 425             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 426             vdwjidx0A        = 2*vdwtype[jnrA+0];
 427
 428             fjx0             = _fjsp_setzero_v2r8();
 429             fjy0             = _fjsp_setzero_v2r8();
 430             fjz0             = _fjsp_setzero_v2r8();
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 437
 438             /* Compute parameters for interactions between i and j atoms */
 439             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 440             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 441
 442             /* Calculate table index by multiplying r with table scale and truncate to integer */
 443             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 444             itab_tmp         = _fjsp_dtox_v2r8(rt);
 445             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 446             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 447             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 448
 449             vfconv.i[0]     *= 12;
 450             vfconv.i[1]     *= 12;
 451
 452             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 453             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 454             F                = _fjsp_setzero_v2r8();
 455             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 456             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 457             H                = _fjsp_setzero_v2r8();
 458             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 459             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 460             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 461             velec            = _fjsp_mul_v2r8(qq00,VV);
 462             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 463             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 464
 465             /* CUBIC SPLINE TABLE DISPERSION */
 466             vfconv.i[0]       += 4;
 467             vfconv.i[1]       += 4;
 468             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 469             F                = _fjsp_setzero_v2r8();
 470             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 471             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 472             H                = _fjsp_setzero_v2r8();
 473             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 474             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 475             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 476             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 477             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 478             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 479
 480             /* CUBIC SPLINE TABLE REPULSION */
 481             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 482             F                = _fjsp_setzero_v2r8();
 483             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 484             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 485             H                = _fjsp_setzero_v2r8();
 486             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 487             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 488             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 489             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 490             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 491             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 492             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 493             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 494
 495             /* Update potential sum for this i atom from the interaction with this j atom. */
 496             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 497             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 498             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 499             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 500
 501             fscal            = _fjsp_add_v2r8(felec,fvdw);
 502
 503             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 504
 505             /* Update vectorial force */
 506             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 507             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 508             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 509
 510             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 511             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 512             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 513
 514             /**************************
 515              * CALCULATE INTERACTIONS *
 516              **************************/
 517
 518             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 519
 520             /* Compute parameters for interactions between i and j atoms */
 521             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 522
 523             /* Calculate table index by multiplying r with table scale and truncate to integer */
 524             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 525             itab_tmp         = _fjsp_dtox_v2r8(rt);
 526             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 527             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 528             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 529
 530             vfconv.i[0]     *= 12;
 531             vfconv.i[1]     *= 12;
 532
 533             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 534             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 535             F                = _fjsp_setzero_v2r8();
 536             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 537             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 538             H                = _fjsp_setzero_v2r8();
 539             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 540             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 541             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 542             velec            = _fjsp_mul_v2r8(qq10,VV);
 543             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 544             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 545
 546             /* Update potential sum for this i atom from the interaction with this j atom. */
 547             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 548             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 549
 550             fscal            = felec;
 551
 552             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 553
 554             /* Update vectorial force */
 555             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 556             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 557             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 558
 559             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 560             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 561             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 562
 563             /**************************
 564              * CALCULATE INTERACTIONS *
 565              **************************/
 566
 567             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 568
 569             /* Compute parameters for interactions between i and j atoms */
 570             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 571
 572             /* Calculate table index by multiplying r with table scale and truncate to integer */
 573             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 574             itab_tmp         = _fjsp_dtox_v2r8(rt);
 575             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 576             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 577             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 578
 579             vfconv.i[0]     *= 12;
 580             vfconv.i[1]     *= 12;
 581
 582             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 583             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 584             F                = _fjsp_setzero_v2r8();
 585             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 586             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 587             H                = _fjsp_setzero_v2r8();
 588             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 589             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 590             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 591             velec            = _fjsp_mul_v2r8(qq20,VV);
 592             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 593             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 594
 595             /* Update potential sum for this i atom from the interaction with this j atom. */
 596             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 597             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 598
 599             fscal            = felec;
 600
 601             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 602
 603             /* Update vectorial force */
 604             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 605             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 606             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 607
 608             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 609             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 610             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 611
 612             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 613
 614             /* Inner loop uses 171 flops */
 615         }
 616
 617         /* End of innermost loop */
 618
 619         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 620                                               f+i_coord_offset,fshift+i_shift_offset);
 621
 622         ggid                        = gid[iidx];
 623         /* Update potential energies */
 624         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 625         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 626
 627         /* Increment number of inner iterations */
 628         inneriter                  += j_index_end - j_index_start;
 629
 630         /* Outer loop uses 20 flops */
 631     }
 632
 633     /* Increment number of outer iterations */
 634     outeriter        += nri;
 635
 636     /* Update outer/inner flops */
 637
 638     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*171);
 639 }
 640 /*
 641  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 642  * Electrostatics interaction: CubicSplineTable
 643  * VdW interaction:            CubicSplineTable
 644  * Geometry:                   Water3-Particle
 645  * Calculate force/pot:        Force
 646  */
 647 void
 648 nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 649                     (t_nblist                    * gmx_restrict       nlist,
 650                      rvec                        * gmx_restrict          xx,
 651                      rvec                        * gmx_restrict          ff,
 652                      t_forcerec                  * gmx_restrict          fr,
 653                      t_mdatoms                   * gmx_restrict     mdatoms,
 654                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 655                      t_nrnb                      * gmx_restrict        nrnb)
 656 {
 657     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 658      * just 0 for non-waters.
 659      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 660      * jnr indices corresponding to data put in the four positions in the SIMD register.
 661      */
 662     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 663     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 664     int              jnrA,jnrB;
 665     int              j_coord_offsetA,j_coord_offsetB;
 666     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 667     real             rcutoff_scalar;
 668     real             *shiftvec,*fshift,*x,*f;
 669     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 670     int              vdwioffset0;
 671     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 672     int              vdwioffset1;
 673     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 674     int              vdwioffset2;
 675     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 676     int              vdwjidx0A,vdwjidx0B;
 677     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 678     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 679     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 680     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 681     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 682     real             *charge;
 683     int              nvdwtype;
 684     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 685     int              *vdwtype;
 686     real             *vdwparam;
 687     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 688     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 689     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 690     real             *vftab;
 691     _fjsp_v2r8       itab_tmp;
 692     _fjsp_v2r8       dummy_mask,cutoff_mask;
 693     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 694     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 695     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 696
 697     x                = xx[0];
 698     f                = ff[0];
 699
 700     nri              = nlist->nri;
 701     iinr             = nlist->iinr;
 702     jindex           = nlist->jindex;
 703     jjnr             = nlist->jjnr;
 704     shiftidx         = nlist->shift;
 705     gid              = nlist->gid;
 706     shiftvec         = fr->shift_vec[0];
 707     fshift           = fr->fshift[0];
 708     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 709     charge           = mdatoms->chargeA;
 710     nvdwtype         = fr->ntype;
 711     vdwparam         = fr->nbfp;
 712     vdwtype          = mdatoms->typeA;
 713
 714     vftab            = kernel_data->table_elec_vdw->data;
 715     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 716
 717     /* Setup water-specific parameters */
 718     inr              = nlist->iinr[0];
 719     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 720     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 721     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 722     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 723
 724     /* Avoid stupid compiler warnings */
 725     jnrA = jnrB = 0;
 726     j_coord_offsetA = 0;
 727     j_coord_offsetB = 0;
 728
 729     outeriter        = 0;
 730     inneriter        = 0;
 731
 732     /* Start outer loop over neighborlists */
 733     for(iidx=0; iidx<nri; iidx++)
 734     {
 735         /* Load shift vector for this list */
 736         i_shift_offset   = DIM*shiftidx[iidx];
 737
 738         /* Load limits for loop over neighbors */
 739         j_index_start    = jindex[iidx];
 740         j_index_end      = jindex[iidx+1];
 741
 742         /* Get outer coordinate index */
 743         inr              = iinr[iidx];
 744         i_coord_offset   = DIM*inr;
 745
 746         /* Load i particle coords and add shift vector */
 747         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 748                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 749
 750         fix0             = _fjsp_setzero_v2r8();
 751         fiy0             = _fjsp_setzero_v2r8();
 752         fiz0             = _fjsp_setzero_v2r8();
 753         fix1             = _fjsp_setzero_v2r8();
 754         fiy1             = _fjsp_setzero_v2r8();
 755         fiz1             = _fjsp_setzero_v2r8();
 756         fix2             = _fjsp_setzero_v2r8();
 757         fiy2             = _fjsp_setzero_v2r8();
 758         fiz2             = _fjsp_setzero_v2r8();
 759
 760         /* Start inner kernel loop */
 761         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 762         {
 763
 764             /* Get j neighbor index, and coordinate index */
 765             jnrA             = jjnr[jidx];
 766             jnrB             = jjnr[jidx+1];
 767             j_coord_offsetA  = DIM*jnrA;
 768             j_coord_offsetB  = DIM*jnrB;
 769
 770             /* load j atom coordinates */
 771             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 772                                               &jx0,&jy0,&jz0);
 773
 774             /* Calculate displacement vector */
 775             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 776             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 777             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 778             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 779             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 780             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 781             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 782             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 783             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 784
 785             /* Calculate squared distance and things based on it */
 786             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 787             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 788             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 789
 790             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 791             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 792             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 793
 794             /* Load parameters for j particles */
 795             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 796             vdwjidx0A        = 2*vdwtype[jnrA+0];
 797             vdwjidx0B        = 2*vdwtype[jnrB+0];
 798
 799             fjx0             = _fjsp_setzero_v2r8();
 800             fjy0             = _fjsp_setzero_v2r8();
 801             fjz0             = _fjsp_setzero_v2r8();
 802
 803             /**************************
 804              * CALCULATE INTERACTIONS *
 805              **************************/
 806
 807             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 808
 809             /* Compute parameters for interactions between i and j atoms */
 810             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 811             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 812                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 813
 814             /* Calculate table index by multiplying r with table scale and truncate to integer */
 815             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 816             itab_tmp         = _fjsp_dtox_v2r8(rt);
 817             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 818             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 819             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 820
 821             vfconv.i[0]     *= 12;
 822             vfconv.i[1]     *= 12;
 823
 824             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 825             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 826             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 827             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 828             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 829             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 830             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 831             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 832             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 833             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 834
 835             /* CUBIC SPLINE TABLE DISPERSION */
 836             vfconv.i[0]       += 4;
 837             vfconv.i[1]       += 4;
 838             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 839             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 840             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 841             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 842             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 843             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 844             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 845             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 846             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 847
 848             /* CUBIC SPLINE TABLE REPULSION */
 849             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 850             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 851             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 852             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 853             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 854             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 855             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 856             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 857             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 858             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 859
 860             fscal            = _fjsp_add_v2r8(felec,fvdw);
 861
 862             /* Update vectorial force */
 863             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 864             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 865             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 866
 867             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 868             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 869             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 870
 871             /**************************
 872              * CALCULATE INTERACTIONS *
 873              **************************/
 874
 875             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 876
 877             /* Compute parameters for interactions between i and j atoms */
 878             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 879
 880             /* Calculate table index by multiplying r with table scale and truncate to integer */
 881             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 882             itab_tmp         = _fjsp_dtox_v2r8(rt);
 883             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 884             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 885             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 886
 887             vfconv.i[0]     *= 12;
 888             vfconv.i[1]     *= 12;
 889
 890             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 891             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 892             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 893             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 894             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 895             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 896             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 897             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 898             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 899             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 900
 901             fscal            = felec;
 902
 903             /* Update vectorial force */
 904             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 905             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 906             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 907
 908             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 909             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 910             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 911
 912             /**************************
 913              * CALCULATE INTERACTIONS *
 914              **************************/
 915
 916             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 917
 918             /* Compute parameters for interactions between i and j atoms */
 919             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 920
 921             /* Calculate table index by multiplying r with table scale and truncate to integer */
 922             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 923             itab_tmp         = _fjsp_dtox_v2r8(rt);
 924             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 925             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 926             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 927
 928             vfconv.i[0]     *= 12;
 929             vfconv.i[1]     *= 12;
 930
 931             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 932             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 933             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 934             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 935             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 936             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 937             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 938             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 939             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 940             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 941
 942             fscal            = felec;
 943
 944             /* Update vectorial force */
 945             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 946             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 947             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 948
 949             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 950             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 951             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 952
 953             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 954
 955             /* Inner loop uses 151 flops */
 956         }
 957
 958         if(jidx<j_index_end)
 959         {
 960
 961             jnrA             = jjnr[jidx];
 962             j_coord_offsetA  = DIM*jnrA;
 963
 964             /* load j atom coordinates */
 965             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 966                                               &jx0,&jy0,&jz0);
 967
 968             /* Calculate displacement vector */
 969             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 970             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 971             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 972             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 973             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 974             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 975             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 976             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 977             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 978
 979             /* Calculate squared distance and things based on it */
 980             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 981             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 982             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 983
 984             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 985             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 986             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 987
 988             /* Load parameters for j particles */
 989             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 990             vdwjidx0A        = 2*vdwtype[jnrA+0];
 991
 992             fjx0             = _fjsp_setzero_v2r8();
 993             fjy0             = _fjsp_setzero_v2r8();
 994             fjz0             = _fjsp_setzero_v2r8();
 995
 996             /**************************
 997              * CALCULATE INTERACTIONS *
 998              **************************/
 999
1000             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1001
1002             /* Compute parameters for interactions between i and j atoms */
1003             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1004             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1005
1006             /* Calculate table index by multiplying r with table scale and truncate to integer */
1007             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1008             itab_tmp         = _fjsp_dtox_v2r8(rt);
1009             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1010             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1011             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1012
1013             vfconv.i[0]     *= 12;
1014             vfconv.i[1]     *= 12;
1015
1016             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1017             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1018             F                = _fjsp_setzero_v2r8();
1019             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1020             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1021             H                = _fjsp_setzero_v2r8();
1022             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1023             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1024             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1025             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
1026
1027             /* CUBIC SPLINE TABLE DISPERSION */
1028             vfconv.i[0]       += 4;
1029             vfconv.i[1]       += 4;
1030             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1031             F                = _fjsp_setzero_v2r8();
1032             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1033             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1034             H                = _fjsp_setzero_v2r8();
1035             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1036             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1037             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1038             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1039
1040             /* CUBIC SPLINE TABLE REPULSION */
1041             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1042             F                = _fjsp_setzero_v2r8();
1043             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1044             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1045             H                = _fjsp_setzero_v2r8();
1046             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1047             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1048             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1049             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1050             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1051
1052             fscal            = _fjsp_add_v2r8(felec,fvdw);
1053
1054             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1055
1056             /* Update vectorial force */
1057             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1058             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1059             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1060
1061             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1062             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1063             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1064
1065             /**************************
1066              * CALCULATE INTERACTIONS *
1067              **************************/
1068
1069             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1070
1071             /* Compute parameters for interactions between i and j atoms */
1072             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1073
1074             /* Calculate table index by multiplying r with table scale and truncate to integer */
1075             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1076             itab_tmp         = _fjsp_dtox_v2r8(rt);
1077             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1078             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1079             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1080
1081             vfconv.i[0]     *= 12;
1082             vfconv.i[1]     *= 12;
1083
1084             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1085             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1086             F                = _fjsp_setzero_v2r8();
1087             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1088             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1089             H                = _fjsp_setzero_v2r8();
1090             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1091             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1092             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1093             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1094
1095             fscal            = felec;
1096
1097             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1098
1099             /* Update vectorial force */
1100             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1101             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1102             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1103
1104             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1105             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1106             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1107
1108             /**************************
1109              * CALCULATE INTERACTIONS *
1110              **************************/
1111
1112             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1113
1114             /* Compute parameters for interactions between i and j atoms */
1115             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1116
1117             /* Calculate table index by multiplying r with table scale and truncate to integer */
1118             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1119             itab_tmp         = _fjsp_dtox_v2r8(rt);
1120             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1121             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1122             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1123
1124             vfconv.i[0]     *= 12;
1125             vfconv.i[1]     *= 12;
1126
1127             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1128             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1129             F                = _fjsp_setzero_v2r8();
1130             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1131             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1132             H                = _fjsp_setzero_v2r8();
1133             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1134             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1135             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1136             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1137
1138             fscal            = felec;
1139
1140             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1141
1142             /* Update vectorial force */
1143             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1144             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1145             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1146
1147             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1148             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1149             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1150
1151             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1152
1153             /* Inner loop uses 151 flops */
1154         }
1155
1156         /* End of innermost loop */
1157
1158         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1159                                               f+i_coord_offset,fshift+i_shift_offset);
1160
1161         /* Increment number of inner iterations */
1162         inneriter                  += j_index_end - j_index_start;
1163
1164         /* Outer loop uses 18 flops */
1165     }
1166
1167     /* Increment number of outer iterations */
1168     outeriter        += nri;
1169
1170     /* Update outer/inner flops */
1171
1172     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*151);
1173 }