src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_VdwLJ_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->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->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             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 210
 211             /* Load parameters for j particles */
 212             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 213             vdwjidx0A        = 2*vdwtype[jnrA+0];
 214             vdwjidx0B        = 2*vdwtype[jnrB+0];
 215
 216             fjx0             = _fjsp_setzero_v2r8();
 217             fjy0             = _fjsp_setzero_v2r8();
 218             fjz0             = _fjsp_setzero_v2r8();
 219
 220             /**************************
 221              * CALCULATE INTERACTIONS *
 222              **************************/
 223
 224             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 225
 226             /* Compute parameters for interactions between i and j atoms */
 227             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 228             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 229                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 230
 231             /* Calculate table index by multiplying r with table scale and truncate to integer */
 232             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 233             itab_tmp         = _fjsp_dtox_v2r8(rt);
 234             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 235             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 236             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 237
 238             vfconv.i[0]     *= 4;
 239             vfconv.i[1]     *= 4;
 240
 241             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 242             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 243             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 244             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 245             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 246             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 247             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 248             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 249             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 250             velec            = _fjsp_mul_v2r8(qq00,VV);
 251             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 252             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 253
 254             /* LENNARD-JONES DISPERSION/REPULSION */
 255
 256             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 257             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 258             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 259             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 260             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 261
 262             /* Update potential sum for this i atom from the interaction with this j atom. */
 263             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 264             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 265
 266             fscal            = _fjsp_add_v2r8(felec,fvdw);
 267
 268             /* Update vectorial force */
 269             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 270             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 271             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 272
 273             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 274             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 275             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 276
 277             /**************************
 278              * CALCULATE INTERACTIONS *
 279              **************************/
 280
 281             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 282
 283             /* Compute parameters for interactions between i and j atoms */
 284             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 285
 286             /* Calculate table index by multiplying r with table scale and truncate to integer */
 287             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 288             itab_tmp         = _fjsp_dtox_v2r8(rt);
 289             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 290             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 291             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 292
 293             vfconv.i[0]     *= 4;
 294             vfconv.i[1]     *= 4;
 295
 296             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 297             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 298             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 299             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 300             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 301             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 302             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 303             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 304             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 305             velec            = _fjsp_mul_v2r8(qq10,VV);
 306             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 307             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 308
 309             /* Update potential sum for this i atom from the interaction with this j atom. */
 310             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 311
 312             fscal            = felec;
 313
 314             /* Update vectorial force */
 315             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 316             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 317             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 318
 319             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 320             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 321             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 322
 323             /**************************
 324              * CALCULATE INTERACTIONS *
 325              **************************/
 326
 327             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 328
 329             /* Compute parameters for interactions between i and j atoms */
 330             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 331
 332             /* Calculate table index by multiplying r with table scale and truncate to integer */
 333             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 334             itab_tmp         = _fjsp_dtox_v2r8(rt);
 335             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 336             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 337             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 338
 339             vfconv.i[0]     *= 4;
 340             vfconv.i[1]     *= 4;
 341
 342             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 343             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 344             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 345             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 346             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 347             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 348             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 349             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 350             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 351             velec            = _fjsp_mul_v2r8(qq20,VV);
 352             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 353             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 354
 355             /* Update potential sum for this i atom from the interaction with this j atom. */
 356             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 357
 358             fscal            = felec;
 359
 360             /* Update vectorial force */
 361             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 362             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 363             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 364
 365             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 366             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 367             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 368
 369             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 370
 371             /* Inner loop uses 154 flops */
 372         }
 373
 374         if(jidx<j_index_end)
 375         {
 376
 377             jnrA             = jjnr[jidx];
 378             j_coord_offsetA  = DIM*jnrA;
 379
 380             /* load j atom coordinates */
 381             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 382                                               &jx0,&jy0,&jz0);
 383
 384             /* Calculate displacement vector */
 385             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 386             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 387             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 388             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 389             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 390             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 391             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 392             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 393             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 394
 395             /* Calculate squared distance and things based on it */
 396             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 397             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 398             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 399
 400             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 401             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 402             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 403
 404             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 405
 406             /* Load parameters for j particles */
 407             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 408             vdwjidx0A        = 2*vdwtype[jnrA+0];
 409
 410             fjx0             = _fjsp_setzero_v2r8();
 411             fjy0             = _fjsp_setzero_v2r8();
 412             fjz0             = _fjsp_setzero_v2r8();
 413
 414             /**************************
 415              * CALCULATE INTERACTIONS *
 416              **************************/
 417
 418             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 419
 420             /* Compute parameters for interactions between i and j atoms */
 421             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 422             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 423
 424             /* Calculate table index by multiplying r with table scale and truncate to integer */
 425             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 426             itab_tmp         = _fjsp_dtox_v2r8(rt);
 427             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 428             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 429             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 430
 431             vfconv.i[0]     *= 4;
 432             vfconv.i[1]     *= 4;
 433
 434             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 435             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 436             F                = _fjsp_setzero_v2r8();
 437             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 438             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 439             H                = _fjsp_setzero_v2r8();
 440             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 441             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 442             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 443             velec            = _fjsp_mul_v2r8(qq00,VV);
 444             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 445             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 446
 447             /* LENNARD-JONES DISPERSION/REPULSION */
 448
 449             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 450             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 451             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 452             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 453             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 454
 455             /* Update potential sum for this i atom from the interaction with this j atom. */
 456             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 457             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 458             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 459             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 460
 461             fscal            = _fjsp_add_v2r8(felec,fvdw);
 462
 463             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 464
 465             /* Update vectorial force */
 466             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 467             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 468             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 469
 470             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 471             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 472             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 473
 474             /**************************
 475              * CALCULATE INTERACTIONS *
 476              **************************/
 477
 478             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 479
 480             /* Compute parameters for interactions between i and j atoms */
 481             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 482
 483             /* Calculate table index by multiplying r with table scale and truncate to integer */
 484             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 485             itab_tmp         = _fjsp_dtox_v2r8(rt);
 486             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 487             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 488             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 489
 490             vfconv.i[0]     *= 4;
 491             vfconv.i[1]     *= 4;
 492
 493             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 494             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 495             F                = _fjsp_setzero_v2r8();
 496             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 497             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 498             H                = _fjsp_setzero_v2r8();
 499             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 500             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 501             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 502             velec            = _fjsp_mul_v2r8(qq10,VV);
 503             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 504             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 505
 506             /* Update potential sum for this i atom from the interaction with this j atom. */
 507             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 508             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 509
 510             fscal            = felec;
 511
 512             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 513
 514             /* Update vectorial force */
 515             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 516             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 517             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 518
 519             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 520             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 521             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 522
 523             /**************************
 524              * CALCULATE INTERACTIONS *
 525              **************************/
 526
 527             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 528
 529             /* Compute parameters for interactions between i and j atoms */
 530             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 531
 532             /* Calculate table index by multiplying r with table scale and truncate to integer */
 533             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 534             itab_tmp         = _fjsp_dtox_v2r8(rt);
 535             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 536             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 537             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 538
 539             vfconv.i[0]     *= 4;
 540             vfconv.i[1]     *= 4;
 541
 542             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 543             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 544             F                = _fjsp_setzero_v2r8();
 545             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 546             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 547             H                = _fjsp_setzero_v2r8();
 548             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 549             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 550             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 551             velec            = _fjsp_mul_v2r8(qq20,VV);
 552             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 553             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 557             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 558
 559             fscal            = felec;
 560
 561             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 562
 563             /* Update vectorial force */
 564             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 565             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 566             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 567
 568             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 569             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 570             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 571
 572             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 573
 574             /* Inner loop uses 154 flops */
 575         }
 576
 577         /* End of innermost loop */
 578
 579         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 580                                               f+i_coord_offset,fshift+i_shift_offset);
 581
 582         ggid                        = gid[iidx];
 583         /* Update potential energies */
 584         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 585         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 586
 587         /* Increment number of inner iterations */
 588         inneriter                  += j_index_end - j_index_start;
 589
 590         /* Outer loop uses 20 flops */
 591     }
 592
 593     /* Increment number of outer iterations */
 594     outeriter        += nri;
 595
 596     /* Update outer/inner flops */
 597
 598     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 599 }
 600 /*
 601  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 602  * Electrostatics interaction: CubicSplineTable
 603  * VdW interaction:            LennardJones
 604  * Geometry:                   Water3-Particle
 605  * Calculate force/pot:        Force
 606  */
 607 void
 608 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 609                     (t_nblist                    * gmx_restrict       nlist,
 610                      rvec                        * gmx_restrict          xx,
 611                      rvec                        * gmx_restrict          ff,
 612                      t_forcerec                  * gmx_restrict          fr,
 613                      t_mdatoms                   * gmx_restrict     mdatoms,
 614                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 615                      t_nrnb                      * gmx_restrict        nrnb)
 616 {
 617     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 618      * just 0 for non-waters.
 619      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 620      * jnr indices corresponding to data put in the four positions in the SIMD register.
 621      */
 622     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 623     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 624     int              jnrA,jnrB;
 625     int              j_coord_offsetA,j_coord_offsetB;
 626     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 627     real             rcutoff_scalar;
 628     real             *shiftvec,*fshift,*x,*f;
 629     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 630     int              vdwioffset0;
 631     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 632     int              vdwioffset1;
 633     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 634     int              vdwioffset2;
 635     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 636     int              vdwjidx0A,vdwjidx0B;
 637     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 638     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 639     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 640     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 641     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 642     real             *charge;
 643     int              nvdwtype;
 644     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 645     int              *vdwtype;
 646     real             *vdwparam;
 647     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 648     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 649     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 650     real             *vftab;
 651     _fjsp_v2r8       itab_tmp;
 652     _fjsp_v2r8       dummy_mask,cutoff_mask;
 653     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 654     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 655     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 656
 657     x                = xx[0];
 658     f                = ff[0];
 659
 660     nri              = nlist->nri;
 661     iinr             = nlist->iinr;
 662     jindex           = nlist->jindex;
 663     jjnr             = nlist->jjnr;
 664     shiftidx         = nlist->shift;
 665     gid              = nlist->gid;
 666     shiftvec         = fr->shift_vec[0];
 667     fshift           = fr->fshift[0];
 668     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 669     charge           = mdatoms->chargeA;
 670     nvdwtype         = fr->ntype;
 671     vdwparam         = fr->nbfp;
 672     vdwtype          = mdatoms->typeA;
 673
 674     vftab            = kernel_data->table_elec->data;
 675     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 676
 677     /* Setup water-specific parameters */
 678     inr              = nlist->iinr[0];
 679     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 680     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 681     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 682     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 683
 684     /* Avoid stupid compiler warnings */
 685     jnrA = jnrB = 0;
 686     j_coord_offsetA = 0;
 687     j_coord_offsetB = 0;
 688
 689     outeriter        = 0;
 690     inneriter        = 0;
 691
 692     /* Start outer loop over neighborlists */
 693     for(iidx=0; iidx<nri; iidx++)
 694     {
 695         /* Load shift vector for this list */
 696         i_shift_offset   = DIM*shiftidx[iidx];
 697
 698         /* Load limits for loop over neighbors */
 699         j_index_start    = jindex[iidx];
 700         j_index_end      = jindex[iidx+1];
 701
 702         /* Get outer coordinate index */
 703         inr              = iinr[iidx];
 704         i_coord_offset   = DIM*inr;
 705
 706         /* Load i particle coords and add shift vector */
 707         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 708                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 709
 710         fix0             = _fjsp_setzero_v2r8();
 711         fiy0             = _fjsp_setzero_v2r8();
 712         fiz0             = _fjsp_setzero_v2r8();
 713         fix1             = _fjsp_setzero_v2r8();
 714         fiy1             = _fjsp_setzero_v2r8();
 715         fiz1             = _fjsp_setzero_v2r8();
 716         fix2             = _fjsp_setzero_v2r8();
 717         fiy2             = _fjsp_setzero_v2r8();
 718         fiz2             = _fjsp_setzero_v2r8();
 719
 720         /* Start inner kernel loop */
 721         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 722         {
 723
 724             /* Get j neighbor index, and coordinate index */
 725             jnrA             = jjnr[jidx];
 726             jnrB             = jjnr[jidx+1];
 727             j_coord_offsetA  = DIM*jnrA;
 728             j_coord_offsetB  = DIM*jnrB;
 729
 730             /* load j atom coordinates */
 731             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 732                                               &jx0,&jy0,&jz0);
 733
 734             /* Calculate displacement vector */
 735             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 736             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 737             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 738             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 739             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 740             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 741             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 742             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 743             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 744
 745             /* Calculate squared distance and things based on it */
 746             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 747             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 748             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 749
 750             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 751             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 752             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 753
 754             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 755
 756             /* Load parameters for j particles */
 757             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 758             vdwjidx0A        = 2*vdwtype[jnrA+0];
 759             vdwjidx0B        = 2*vdwtype[jnrB+0];
 760
 761             fjx0             = _fjsp_setzero_v2r8();
 762             fjy0             = _fjsp_setzero_v2r8();
 763             fjz0             = _fjsp_setzero_v2r8();
 764
 765             /**************************
 766              * CALCULATE INTERACTIONS *
 767              **************************/
 768
 769             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 770
 771             /* Compute parameters for interactions between i and j atoms */
 772             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 773             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 774                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 775
 776             /* Calculate table index by multiplying r with table scale and truncate to integer */
 777             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 778             itab_tmp         = _fjsp_dtox_v2r8(rt);
 779             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 780             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 781             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 782
 783             vfconv.i[0]     *= 4;
 784             vfconv.i[1]     *= 4;
 785
 786             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 787             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 788             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 789             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 790             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 791             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 792             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 793             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 794             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 795             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 796
 797             /* LENNARD-JONES DISPERSION/REPULSION */
 798
 799             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 800             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 801
 802             fscal            = _fjsp_add_v2r8(felec,fvdw);
 803
 804             /* Update vectorial force */
 805             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 806             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 807             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 808
 809             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 810             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 811             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 812
 813             /**************************
 814              * CALCULATE INTERACTIONS *
 815              **************************/
 816
 817             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 818
 819             /* Compute parameters for interactions between i and j atoms */
 820             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 821
 822             /* Calculate table index by multiplying r with table scale and truncate to integer */
 823             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 824             itab_tmp         = _fjsp_dtox_v2r8(rt);
 825             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 826             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 827             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 828
 829             vfconv.i[0]     *= 4;
 830             vfconv.i[1]     *= 4;
 831
 832             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 833             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 834             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 835             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 836             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 837             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 838             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 839             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 840             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 841             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 842
 843             fscal            = felec;
 844
 845             /* Update vectorial force */
 846             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 847             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 848             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 849
 850             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 851             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 852             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 859
 860             /* Compute parameters for interactions between i and j atoms */
 861             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 862
 863             /* Calculate table index by multiplying r with table scale and truncate to integer */
 864             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 865             itab_tmp         = _fjsp_dtox_v2r8(rt);
 866             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 867             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 868             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 869
 870             vfconv.i[0]     *= 4;
 871             vfconv.i[1]     *= 4;
 872
 873             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 874             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 875             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 876             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 877             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 878             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 879             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 880             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 881             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 882             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 883
 884             fscal            = felec;
 885
 886             /* Update vectorial force */
 887             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 888             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 889             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 890
 891             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 892             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 893             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 894
 895             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 896
 897             /* Inner loop uses 137 flops */
 898         }
 899
 900         if(jidx<j_index_end)
 901         {
 902
 903             jnrA             = jjnr[jidx];
 904             j_coord_offsetA  = DIM*jnrA;
 905
 906             /* load j atom coordinates */
 907             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 908                                               &jx0,&jy0,&jz0);
 909
 910             /* Calculate displacement vector */
 911             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 912             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 913             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 914             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 915             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 916             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 917             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 918             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 919             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 920
 921             /* Calculate squared distance and things based on it */
 922             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 923             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 924             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 925
 926             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 927             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 928             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 929
 930             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 931
 932             /* Load parameters for j particles */
 933             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 934             vdwjidx0A        = 2*vdwtype[jnrA+0];
 935
 936             fjx0             = _fjsp_setzero_v2r8();
 937             fjy0             = _fjsp_setzero_v2r8();
 938             fjz0             = _fjsp_setzero_v2r8();
 939
 940             /**************************
 941              * CALCULATE INTERACTIONS *
 942              **************************/
 943
 944             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 945
 946             /* Compute parameters for interactions between i and j atoms */
 947             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 948             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 949
 950             /* Calculate table index by multiplying r with table scale and truncate to integer */
 951             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 952             itab_tmp         = _fjsp_dtox_v2r8(rt);
 953             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 954             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 955             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 956
 957             vfconv.i[0]     *= 4;
 958             vfconv.i[1]     *= 4;
 959
 960             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 961             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 962             F                = _fjsp_setzero_v2r8();
 963             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 964             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 965             H                = _fjsp_setzero_v2r8();
 966             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 967             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 968             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 969             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 970
 971             /* LENNARD-JONES DISPERSION/REPULSION */
 972
 973             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 974             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 975
 976             fscal            = _fjsp_add_v2r8(felec,fvdw);
 977
 978             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 979
 980             /* Update vectorial force */
 981             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 982             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 983             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 984
 985             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 986             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 987             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 988
 989             /**************************
 990              * CALCULATE INTERACTIONS *
 991              **************************/
 992
 993             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 994
 995             /* Compute parameters for interactions between i and j atoms */
 996             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 997
 998             /* Calculate table index by multiplying r with table scale and truncate to integer */
 999             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1000             itab_tmp         = _fjsp_dtox_v2r8(rt);
1001             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1002             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1003             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1004
1005             vfconv.i[0]     *= 4;
1006             vfconv.i[1]     *= 4;
1007
1008             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1009             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1010             F                = _fjsp_setzero_v2r8();
1011             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1012             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1013             H                = _fjsp_setzero_v2r8();
1014             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1015             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1016             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1017             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1018
1019             fscal            = felec;
1020
1021             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1022
1023             /* Update vectorial force */
1024             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1025             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1026             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1027
1028             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1029             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1030             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1031
1032             /**************************
1033              * CALCULATE INTERACTIONS *
1034              **************************/
1035
1036             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1037
1038             /* Compute parameters for interactions between i and j atoms */
1039             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1040
1041             /* Calculate table index by multiplying r with table scale and truncate to integer */
1042             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1043             itab_tmp         = _fjsp_dtox_v2r8(rt);
1044             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1045             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1046             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1047
1048             vfconv.i[0]     *= 4;
1049             vfconv.i[1]     *= 4;
1050
1051             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1052             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1053             F                = _fjsp_setzero_v2r8();
1054             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1055             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1056             H                = _fjsp_setzero_v2r8();
1057             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1058             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1059             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1060             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1061
1062             fscal            = felec;
1063
1064             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1065
1066             /* Update vectorial force */
1067             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1068             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1069             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1070
1071             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1072             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1073             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1074
1075             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1076
1077             /* Inner loop uses 137 flops */
1078         }
1079
1080         /* End of innermost loop */
1081
1082         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1083                                               f+i_coord_offset,fshift+i_shift_offset);
1084
1085         /* Increment number of inner iterations */
1086         inneriter                  += j_index_end - j_index_start;
1087
1088         /* Outer loop uses 18 flops */
1089     }
1090
1091     /* Increment number of outer iterations */
1092     outeriter        += nri;
1093
1094     /* Update outer/inner flops */
1095
1096     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*137);
1097 }