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 "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: CubicSplineTable
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       itab_tmp;
 101     _fjsp_v2r8       dummy_mask,cutoff_mask;
 102     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 103     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 104     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 105
 106     x                = xx[0];
 107     f                = ff[0];
 108
 109     nri              = nlist->nri;
 110     iinr             = nlist->iinr;
 111     jindex           = nlist->jindex;
 112     jjnr             = nlist->jjnr;
 113     shiftidx         = nlist->shift;
 114     gid              = nlist->gid;
 115     shiftvec         = fr->shift_vec[0];
 116     fshift           = fr->fshift[0];
 117     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 118     charge           = mdatoms->chargeA;
 119     nvdwtype         = fr->ntype;
 120     vdwparam         = fr->nbfp;
 121     vdwtype          = mdatoms->typeA;
 122
 123     vftab            = kernel_data->table_elec->data;
 124     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 125
 126     /* Setup water-specific parameters */
 127     inr              = nlist->iinr[0];
 128     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 129     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 130     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 131     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133     /* Avoid stupid compiler warnings */
 134     jnrA = jnrB = 0;
 135     j_coord_offsetA = 0;
 136     j_coord_offsetB = 0;
 137
 138     outeriter        = 0;
 139     inneriter        = 0;
 140
 141     /* Start outer loop over neighborlists */
 142     for(iidx=0; iidx<nri; iidx++)
 143     {
 144         /* Load shift vector for this list */
 145         i_shift_offset   = DIM*shiftidx[iidx];
 146
 147         /* Load limits for loop over neighbors */
 148         j_index_start    = jindex[iidx];
 149         j_index_end      = jindex[iidx+1];
 150
 151         /* Get outer coordinate index */
 152         inr              = iinr[iidx];
 153         i_coord_offset   = DIM*inr;
 154
 155         /* Load i particle coords and add shift vector */
 156         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 157                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 158
 159         fix0             = _fjsp_setzero_v2r8();
 160         fiy0             = _fjsp_setzero_v2r8();
 161         fiz0             = _fjsp_setzero_v2r8();
 162         fix1             = _fjsp_setzero_v2r8();
 163         fiy1             = _fjsp_setzero_v2r8();
 164         fiz1             = _fjsp_setzero_v2r8();
 165         fix2             = _fjsp_setzero_v2r8();
 166         fiy2             = _fjsp_setzero_v2r8();
 167         fiz2             = _fjsp_setzero_v2r8();
 168
 169         /* Reset potential sums */
 170         velecsum         = _fjsp_setzero_v2r8();
 171         vvdwsum          = _fjsp_setzero_v2r8();
 172
 173         /* Start inner kernel loop */
 174         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 175         {
 176
 177             /* Get j neighbor index, and coordinate index */
 178             jnrA             = jjnr[jidx];
 179             jnrB             = jjnr[jidx+1];
 180             j_coord_offsetA  = DIM*jnrA;
 181             j_coord_offsetB  = DIM*jnrB;
 182
 183             /* load j atom coordinates */
 184             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 185                                               &jx0,&jy0,&jz0);
 186
 187             /* Calculate displacement vector */
 188             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 189             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 190             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 191             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 192             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 193             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 194             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 195             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 196             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 197
 198             /* Calculate squared distance and things based on it */
 199             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 200             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 201             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 202
 203             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 204             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 205             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 206
 207             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 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]     *= 4;
 237             vfconv.i[1]     *= 4;
 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             /* LENNARD-JONES DISPERSION/REPULSION */
 253
 254             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 255             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 256             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 257             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 258             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 259
 260             /* Update potential sum for this i atom from the interaction with this j atom. */
 261             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 262             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 263
 264             fscal            = _fjsp_add_v2r8(felec,fvdw);
 265
 266             /* Update vectorial force */
 267             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 268             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 269             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 270
 271             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 272             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 273             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 274
 275             /**************************
 276              * CALCULATE INTERACTIONS *
 277              **************************/
 278
 279             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 280
 281             /* Compute parameters for interactions between i and j atoms */
 282             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 283
 284             /* Calculate table index by multiplying r with table scale and truncate to integer */
 285             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 286             itab_tmp         = _fjsp_dtox_v2r8(rt);
 287             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 288             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 289             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 290
 291             vfconv.i[0]     *= 4;
 292             vfconv.i[1]     *= 4;
 293
 294             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 295             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 296             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 297             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 298             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 299             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 300             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 301             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 302             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 303             velec            = _fjsp_mul_v2r8(qq10,VV);
 304             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 305             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 306
 307             /* Update potential sum for this i atom from the interaction with this j atom. */
 308             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 309
 310             fscal            = felec;
 311
 312             /* Update vectorial force */
 313             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 314             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 315             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 316
 317             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 318             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 319             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 320
 321             /**************************
 322              * CALCULATE INTERACTIONS *
 323              **************************/
 324
 325             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 326
 327             /* Compute parameters for interactions between i and j atoms */
 328             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 329
 330             /* Calculate table index by multiplying r with table scale and truncate to integer */
 331             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 332             itab_tmp         = _fjsp_dtox_v2r8(rt);
 333             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 334             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 335             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 336
 337             vfconv.i[0]     *= 4;
 338             vfconv.i[1]     *= 4;
 339
 340             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 341             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 342             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 343             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 344             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 345             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 346             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 347             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 348             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 349             velec            = _fjsp_mul_v2r8(qq20,VV);
 350             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 351             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 352
 353             /* Update potential sum for this i atom from the interaction with this j atom. */
 354             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 355
 356             fscal            = felec;
 357
 358             /* Update vectorial force */
 359             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 360             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 361             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 362
 363             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 364             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 365             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 366
 367             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 368
 369             /* Inner loop uses 154 flops */
 370         }
 371
 372         if(jidx<j_index_end)
 373         {
 374
 375             jnrA             = jjnr[jidx];
 376             j_coord_offsetA  = DIM*jnrA;
 377
 378             /* load j atom coordinates */
 379             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 380                                               &jx0,&jy0,&jz0);
 381
 382             /* Calculate displacement vector */
 383             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 384             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 385             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 386             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 387             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 388             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 389             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 390             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 391             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 392
 393             /* Calculate squared distance and things based on it */
 394             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 395             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 396             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 397
 398             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 399             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 400             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 401
 402             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 403
 404             /* Load parameters for j particles */
 405             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 406             vdwjidx0A        = 2*vdwtype[jnrA+0];
 407
 408             fjx0             = _fjsp_setzero_v2r8();
 409             fjy0             = _fjsp_setzero_v2r8();
 410             fjz0             = _fjsp_setzero_v2r8();
 411
 412             /**************************
 413              * CALCULATE INTERACTIONS *
 414              **************************/
 415
 416             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 417
 418             /* Compute parameters for interactions between i and j atoms */
 419             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 420             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 421
 422             /* Calculate table index by multiplying r with table scale and truncate to integer */
 423             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 424             itab_tmp         = _fjsp_dtox_v2r8(rt);
 425             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 426             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 427             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 428
 429             vfconv.i[0]     *= 4;
 430             vfconv.i[1]     *= 4;
 431
 432             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 433             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 434             F                = _fjsp_setzero_v2r8();
 435             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 436             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 437             H                = _fjsp_setzero_v2r8();
 438             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 439             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 440             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 441             velec            = _fjsp_mul_v2r8(qq00,VV);
 442             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 443             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 444
 445             /* LENNARD-JONES DISPERSION/REPULSION */
 446
 447             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 448             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 449             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 450             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 451             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 452
 453             /* Update potential sum for this i atom from the interaction with this j atom. */
 454             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 455             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 456             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 457             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 458
 459             fscal            = _fjsp_add_v2r8(felec,fvdw);
 460
 461             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 462
 463             /* Update vectorial force */
 464             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 465             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 466             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 467
 468             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 469             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 470             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 471
 472             /**************************
 473              * CALCULATE INTERACTIONS *
 474              **************************/
 475
 476             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 477
 478             /* Compute parameters for interactions between i and j atoms */
 479             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 480
 481             /* Calculate table index by multiplying r with table scale and truncate to integer */
 482             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 483             itab_tmp         = _fjsp_dtox_v2r8(rt);
 484             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 485             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 486             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 487
 488             vfconv.i[0]     *= 4;
 489             vfconv.i[1]     *= 4;
 490
 491             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 492             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 493             F                = _fjsp_setzero_v2r8();
 494             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 495             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 496             H                = _fjsp_setzero_v2r8();
 497             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 498             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 499             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 500             velec            = _fjsp_mul_v2r8(qq10,VV);
 501             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 502             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 503
 504             /* Update potential sum for this i atom from the interaction with this j atom. */
 505             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 506             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 507
 508             fscal            = felec;
 509
 510             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 511
 512             /* Update vectorial force */
 513             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 514             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 515             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 516
 517             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 518             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 519             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 520
 521             /**************************
 522              * CALCULATE INTERACTIONS *
 523              **************************/
 524
 525             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 526
 527             /* Compute parameters for interactions between i and j atoms */
 528             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 529
 530             /* Calculate table index by multiplying r with table scale and truncate to integer */
 531             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 532             itab_tmp         = _fjsp_dtox_v2r8(rt);
 533             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 534             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 535             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 536
 537             vfconv.i[0]     *= 4;
 538             vfconv.i[1]     *= 4;
 539
 540             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 541             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 542             F                = _fjsp_setzero_v2r8();
 543             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 544             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 545             H                = _fjsp_setzero_v2r8();
 546             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 547             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 548             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 549             velec            = _fjsp_mul_v2r8(qq20,VV);
 550             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 551             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 552
 553             /* Update potential sum for this i atom from the interaction with this j atom. */
 554             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 555             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 556
 557             fscal            = felec;
 558
 559             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 560
 561             /* Update vectorial force */
 562             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 563             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 564             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 565
 566             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 567             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 568             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 569
 570             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 571
 572             /* Inner loop uses 154 flops */
 573         }
 574
 575         /* End of innermost loop */
 576
 577         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 578                                               f+i_coord_offset,fshift+i_shift_offset);
 579
 580         ggid                        = gid[iidx];
 581         /* Update potential energies */
 582         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 583         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 584
 585         /* Increment number of inner iterations */
 586         inneriter                  += j_index_end - j_index_start;
 587
 588         /* Outer loop uses 20 flops */
 589     }
 590
 591     /* Increment number of outer iterations */
 592     outeriter        += nri;
 593
 594     /* Update outer/inner flops */
 595
 596     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*154);
 597 }
 598 /*
 599  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 600  * Electrostatics interaction: CubicSplineTable
 601  * VdW interaction:            LennardJones
 602  * Geometry:                   Water3-Particle
 603  * Calculate force/pot:        Force
 604  */
 605 void
 606 nb_kernel_ElecCSTab_VdwLJ_GeomW3P1_F_sparc64_hpc_ace_double
 607                     (t_nblist                    * gmx_restrict       nlist,
 608                      rvec                        * gmx_restrict          xx,
 609                      rvec                        * gmx_restrict          ff,
 610                      t_forcerec                  * gmx_restrict          fr,
 611                      t_mdatoms                   * gmx_restrict     mdatoms,
 612                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 613                      t_nrnb                      * gmx_restrict        nrnb)
 614 {
 615     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 616      * just 0 for non-waters.
 617      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 618      * jnr indices corresponding to data put in the four positions in the SIMD register.
 619      */
 620     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 621     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 622     int              jnrA,jnrB;
 623     int              j_coord_offsetA,j_coord_offsetB;
 624     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 625     real             rcutoff_scalar;
 626     real             *shiftvec,*fshift,*x,*f;
 627     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 628     int              vdwioffset0;
 629     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 630     int              vdwioffset1;
 631     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 632     int              vdwioffset2;
 633     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 634     int              vdwjidx0A,vdwjidx0B;
 635     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 636     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 637     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 638     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 639     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 640     real             *charge;
 641     int              nvdwtype;
 642     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 643     int              *vdwtype;
 644     real             *vdwparam;
 645     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 646     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 647     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 648     real             *vftab;
 649     _fjsp_v2r8       itab_tmp;
 650     _fjsp_v2r8       dummy_mask,cutoff_mask;
 651     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 652     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 653     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 654
 655     x                = xx[0];
 656     f                = ff[0];
 657
 658     nri              = nlist->nri;
 659     iinr             = nlist->iinr;
 660     jindex           = nlist->jindex;
 661     jjnr             = nlist->jjnr;
 662     shiftidx         = nlist->shift;
 663     gid              = nlist->gid;
 664     shiftvec         = fr->shift_vec[0];
 665     fshift           = fr->fshift[0];
 666     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 667     charge           = mdatoms->chargeA;
 668     nvdwtype         = fr->ntype;
 669     vdwparam         = fr->nbfp;
 670     vdwtype          = mdatoms->typeA;
 671
 672     vftab            = kernel_data->table_elec->data;
 673     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 674
 675     /* Setup water-specific parameters */
 676     inr              = nlist->iinr[0];
 677     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 678     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 679     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 680     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 681
 682     /* Avoid stupid compiler warnings */
 683     jnrA = jnrB = 0;
 684     j_coord_offsetA = 0;
 685     j_coord_offsetB = 0;
 686
 687     outeriter        = 0;
 688     inneriter        = 0;
 689
 690     /* Start outer loop over neighborlists */
 691     for(iidx=0; iidx<nri; iidx++)
 692     {
 693         /* Load shift vector for this list */
 694         i_shift_offset   = DIM*shiftidx[iidx];
 695
 696         /* Load limits for loop over neighbors */
 697         j_index_start    = jindex[iidx];
 698         j_index_end      = jindex[iidx+1];
 699
 700         /* Get outer coordinate index */
 701         inr              = iinr[iidx];
 702         i_coord_offset   = DIM*inr;
 703
 704         /* Load i particle coords and add shift vector */
 705         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 706                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 707
 708         fix0             = _fjsp_setzero_v2r8();
 709         fiy0             = _fjsp_setzero_v2r8();
 710         fiz0             = _fjsp_setzero_v2r8();
 711         fix1             = _fjsp_setzero_v2r8();
 712         fiy1             = _fjsp_setzero_v2r8();
 713         fiz1             = _fjsp_setzero_v2r8();
 714         fix2             = _fjsp_setzero_v2r8();
 715         fiy2             = _fjsp_setzero_v2r8();
 716         fiz2             = _fjsp_setzero_v2r8();
 717
 718         /* Start inner kernel loop */
 719         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 720         {
 721
 722             /* Get j neighbor index, and coordinate index */
 723             jnrA             = jjnr[jidx];
 724             jnrB             = jjnr[jidx+1];
 725             j_coord_offsetA  = DIM*jnrA;
 726             j_coord_offsetB  = DIM*jnrB;
 727
 728             /* load j atom coordinates */
 729             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 730                                               &jx0,&jy0,&jz0);
 731
 732             /* Calculate displacement vector */
 733             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 734             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 735             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 736             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 737             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 738             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 739             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 740             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 741             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 742
 743             /* Calculate squared distance and things based on it */
 744             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 745             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 746             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 747
 748             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 749             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 750             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 751
 752             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 753
 754             /* Load parameters for j particles */
 755             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 756             vdwjidx0A        = 2*vdwtype[jnrA+0];
 757             vdwjidx0B        = 2*vdwtype[jnrB+0];
 758
 759             fjx0             = _fjsp_setzero_v2r8();
 760             fjy0             = _fjsp_setzero_v2r8();
 761             fjz0             = _fjsp_setzero_v2r8();
 762
 763             /**************************
 764              * CALCULATE INTERACTIONS *
 765              **************************/
 766
 767             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 768
 769             /* Compute parameters for interactions between i and j atoms */
 770             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 771             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 772                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 773
 774             /* Calculate table index by multiplying r with table scale and truncate to integer */
 775             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 776             itab_tmp         = _fjsp_dtox_v2r8(rt);
 777             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 778             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 779             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 780
 781             vfconv.i[0]     *= 4;
 782             vfconv.i[1]     *= 4;
 783
 784             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 785             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 786             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 787             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 788             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 789             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 790             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 791             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 792             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 793             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 794
 795             /* LENNARD-JONES DISPERSION/REPULSION */
 796
 797             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 798             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 799
 800             fscal            = _fjsp_add_v2r8(felec,fvdw);
 801
 802             /* Update vectorial force */
 803             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 804             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 805             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 806
 807             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 808             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 809             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 810
 811             /**************************
 812              * CALCULATE INTERACTIONS *
 813              **************************/
 814
 815             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 816
 817             /* Compute parameters for interactions between i and j atoms */
 818             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 819
 820             /* Calculate table index by multiplying r with table scale and truncate to integer */
 821             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 822             itab_tmp         = _fjsp_dtox_v2r8(rt);
 823             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 824             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 825             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 826
 827             vfconv.i[0]     *= 4;
 828             vfconv.i[1]     *= 4;
 829
 830             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 831             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 832             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 833             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 834             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 835             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 836             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 837             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 838             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 839             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 840
 841             fscal            = felec;
 842
 843             /* Update vectorial force */
 844             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 845             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 846             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 847
 848             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 849             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 850             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 851
 852             /**************************
 853              * CALCULATE INTERACTIONS *
 854              **************************/
 855
 856             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 857
 858             /* Compute parameters for interactions between i and j atoms */
 859             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 860
 861             /* Calculate table index by multiplying r with table scale and truncate to integer */
 862             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 863             itab_tmp         = _fjsp_dtox_v2r8(rt);
 864             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 865             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 866             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 867
 868             vfconv.i[0]     *= 4;
 869             vfconv.i[1]     *= 4;
 870
 871             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 872             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 873             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 874             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 875             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 876             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 877             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 878             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 879             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 880             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 881
 882             fscal            = felec;
 883
 884             /* Update vectorial force */
 885             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 886             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 887             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 888
 889             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 890             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 891             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 892
 893             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 894
 895             /* Inner loop uses 137 flops */
 896         }
 897
 898         if(jidx<j_index_end)
 899         {
 900
 901             jnrA             = jjnr[jidx];
 902             j_coord_offsetA  = DIM*jnrA;
 903
 904             /* load j atom coordinates */
 905             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 906                                               &jx0,&jy0,&jz0);
 907
 908             /* Calculate displacement vector */
 909             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 910             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 911             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 912             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 913             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 914             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 915             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 916             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 917             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 918
 919             /* Calculate squared distance and things based on it */
 920             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 921             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 922             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 923
 924             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 925             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 926             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 927
 928             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 929
 930             /* Load parameters for j particles */
 931             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 932             vdwjidx0A        = 2*vdwtype[jnrA+0];
 933
 934             fjx0             = _fjsp_setzero_v2r8();
 935             fjy0             = _fjsp_setzero_v2r8();
 936             fjz0             = _fjsp_setzero_v2r8();
 937
 938             /**************************
 939              * CALCULATE INTERACTIONS *
 940              **************************/
 941
 942             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 943
 944             /* Compute parameters for interactions between i and j atoms */
 945             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 946             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 947
 948             /* Calculate table index by multiplying r with table scale and truncate to integer */
 949             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 950             itab_tmp         = _fjsp_dtox_v2r8(rt);
 951             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 952             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 953             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 954
 955             vfconv.i[0]     *= 4;
 956             vfconv.i[1]     *= 4;
 957
 958             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 959             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 960             F                = _fjsp_setzero_v2r8();
 961             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 962             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 963             H                = _fjsp_setzero_v2r8();
 964             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 965             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 966             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 967             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 968
 969             /* LENNARD-JONES DISPERSION/REPULSION */
 970
 971             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 972             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 973
 974             fscal            = _fjsp_add_v2r8(felec,fvdw);
 975
 976             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 977
 978             /* Update vectorial force */
 979             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 980             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 981             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 982
 983             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 984             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 985             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 986
 987             /**************************
 988              * CALCULATE INTERACTIONS *
 989              **************************/
 990
 991             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 992
 993             /* Compute parameters for interactions between i and j atoms */
 994             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 995
 996             /* Calculate table index by multiplying r with table scale and truncate to integer */
 997             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 998             itab_tmp         = _fjsp_dtox_v2r8(rt);
 999             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1000             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1001             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1002
1003             vfconv.i[0]     *= 4;
1004             vfconv.i[1]     *= 4;
1005
1006             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1007             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1008             F                = _fjsp_setzero_v2r8();
1009             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1010             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1011             H                = _fjsp_setzero_v2r8();
1012             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1013             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1014             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1015             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1016
1017             fscal            = felec;
1018
1019             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1020
1021             /* Update vectorial force */
1022             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1023             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1024             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1025
1026             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1027             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1028             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1029
1030             /**************************
1031              * CALCULATE INTERACTIONS *
1032              **************************/
1033
1034             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1035
1036             /* Compute parameters for interactions between i and j atoms */
1037             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1038
1039             /* Calculate table index by multiplying r with table scale and truncate to integer */
1040             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1041             itab_tmp         = _fjsp_dtox_v2r8(rt);
1042             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1043             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1044             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1045
1046             vfconv.i[0]     *= 4;
1047             vfconv.i[1]     *= 4;
1048
1049             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1050             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1051             F                = _fjsp_setzero_v2r8();
1052             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1053             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1054             H                = _fjsp_setzero_v2r8();
1055             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1056             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1057             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1058             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1059
1060             fscal            = felec;
1061
1062             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1063
1064             /* Update vectorial force */
1065             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1066             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1067             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1068
1069             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1070             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1071             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1072
1073             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1074
1075             /* Inner loop uses 137 flops */
1076         }
1077
1078         /* End of innermost loop */
1079
1080         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1081                                               f+i_coord_offset,fshift+i_shift_offset);
1082
1083         /* Increment number of inner iterations */
1084         inneriter                  += j_index_end - j_index_start;
1085
1086         /* Outer loop uses 18 flops */
1087     }
1088
1089     /* Increment number of outer iterations */
1090     outeriter        += nri;
1091
1092     /* Update outer/inner flops */
1093
1094     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*137);
1095 }