src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, 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_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwioffset1;
  84     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  85     int              vdwioffset2;
  86     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  87     int              vdwjidx0A,vdwjidx0B;
  88     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  89     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  90     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  91     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  92     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  93     real             *charge;
  94     int              nvdwtype;
  95     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  96     int              *vdwtype;
  97     real             *vdwparam;
  98     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  99     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 100     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 101     real             *vftab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_elec_vdw->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 127
 128     /* Setup water-specific parameters */
 129     inr              = nlist->iinr[0];
 130     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 131     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 132     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 133     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 134
 135     /* Avoid stupid compiler warnings */
 136     jnrA = jnrB = 0;
 137     j_coord_offsetA = 0;
 138     j_coord_offsetB = 0;
 139
 140     outeriter        = 0;
 141     inneriter        = 0;
 142
 143     /* Start outer loop over neighborlists */
 144     for(iidx=0; iidx<nri; iidx++)
 145     {
 146         /* Load shift vector for this list */
 147         i_shift_offset   = DIM*shiftidx[iidx];
 148
 149         /* Load limits for loop over neighbors */
 150         j_index_start    = jindex[iidx];
 151         j_index_end      = jindex[iidx+1];
 152
 153         /* Get outer coordinate index */
 154         inr              = iinr[iidx];
 155         i_coord_offset   = DIM*inr;
 156
 157         /* Load i particle coords and add shift vector */
 158         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 159                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 160
 161         fix0             = _fjsp_setzero_v2r8();
 162         fiy0             = _fjsp_setzero_v2r8();
 163         fiz0             = _fjsp_setzero_v2r8();
 164         fix1             = _fjsp_setzero_v2r8();
 165         fiy1             = _fjsp_setzero_v2r8();
 166         fiz1             = _fjsp_setzero_v2r8();
 167         fix2             = _fjsp_setzero_v2r8();
 168         fiy2             = _fjsp_setzero_v2r8();
 169         fiz2             = _fjsp_setzero_v2r8();
 170
 171         /* Reset potential sums */
 172         velecsum         = _fjsp_setzero_v2r8();
 173         vvdwsum          = _fjsp_setzero_v2r8();
 174
 175         /* Start inner kernel loop */
 176         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 177         {
 178
 179             /* Get j neighbor index, and coordinate index */
 180             jnrA             = jjnr[jidx];
 181             jnrB             = jjnr[jidx+1];
 182             j_coord_offsetA  = DIM*jnrA;
 183             j_coord_offsetB  = DIM*jnrB;
 184
 185             /* load j atom coordinates */
 186             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 187                                               &jx0,&jy0,&jz0);
 188
 189             /* Calculate displacement vector */
 190             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 191             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 192             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 193             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 194             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 195             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 196             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 197             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 198             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 199
 200             /* Calculate squared distance and things based on it */
 201             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 202             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 203             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 204
 205             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 206             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 207             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 208
 209             /* Load parameters for j particles */
 210             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 211             vdwjidx0A        = 2*vdwtype[jnrA+0];
 212             vdwjidx0B        = 2*vdwtype[jnrB+0];
 213
 214             fjx0             = _fjsp_setzero_v2r8();
 215             fjy0             = _fjsp_setzero_v2r8();
 216             fjz0             = _fjsp_setzero_v2r8();
 217
 218             /**************************
 219              * CALCULATE INTERACTIONS *
 220              **************************/
 221
 222             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 223
 224             /* Compute parameters for interactions between i and j atoms */
 225             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 226             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 227                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 228
 229             /* Calculate table index by multiplying r with table scale and truncate to integer */
 230             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 231             itab_tmp         = _fjsp_dtox_v2r8(rt);
 232             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 233             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 234             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 235
 236             vfconv.i[0]     *= 12;
 237             vfconv.i[1]     *= 12;
 238
 239             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 240             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 241             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 242             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 243             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 244             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 245             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 246             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 247             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 248             velec            = _fjsp_mul_v2r8(qq00,VV);
 249             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 250             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 251
 252             /* CUBIC SPLINE TABLE DISPERSION */
 253             vfconv.i[0]       += 4;
 254             vfconv.i[1]       += 4;
 255             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 256             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 257             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 258             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 259             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 260             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 261             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 262             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 263             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 264             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 265             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 266
 267             /* CUBIC SPLINE TABLE REPULSION */
 268             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 269             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 270             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 271             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 272             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 273             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 274             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 275             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 276             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 277             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 278             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 279             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 280             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 281
 282             /* Update potential sum for this i atom from the interaction with this j atom. */
 283             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 284             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 285
 286             fscal            = _fjsp_add_v2r8(felec,fvdw);
 287
 288             /* Update vectorial force */
 289             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 290             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 291             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 292
 293             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 294             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 295             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 296
 297             /**************************
 298              * CALCULATE INTERACTIONS *
 299              **************************/
 300
 301             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 302
 303             /* Compute parameters for interactions between i and j atoms */
 304             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 305
 306             /* Calculate table index by multiplying r with table scale and truncate to integer */
 307             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 308             itab_tmp         = _fjsp_dtox_v2r8(rt);
 309             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 310             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 311             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 312
 313             vfconv.i[0]     *= 12;
 314             vfconv.i[1]     *= 12;
 315
 316             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 317             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 318             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 319             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 320             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 321             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 322             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 323             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 324             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 325             velec            = _fjsp_mul_v2r8(qq10,VV);
 326             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 327             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 328
 329             /* Update potential sum for this i atom from the interaction with this j atom. */
 330             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 331
 332             fscal            = felec;
 333
 334             /* Update vectorial force */
 335             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 336             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 337             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 338
 339             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 340             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 341             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 342
 343             /**************************
 344              * CALCULATE INTERACTIONS *
 345              **************************/
 346
 347             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 348
 349             /* Compute parameters for interactions between i and j atoms */
 350             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 351
 352             /* Calculate table index by multiplying r with table scale and truncate to integer */
 353             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 354             itab_tmp         = _fjsp_dtox_v2r8(rt);
 355             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 356             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 357             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 358
 359             vfconv.i[0]     *= 12;
 360             vfconv.i[1]     *= 12;
 361
 362             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 363             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 364             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 365             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 366             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 367             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 368             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 369             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 370             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 371             velec            = _fjsp_mul_v2r8(qq20,VV);
 372             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 373             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 374
 375             /* Update potential sum for this i atom from the interaction with this j atom. */
 376             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 377
 378             fscal            = felec;
 379
 380             /* Update vectorial force */
 381             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 382             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 383             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 384
 385             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 386             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 387             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 388
 389             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 390
 391             /* Inner loop uses 171 flops */
 392         }
 393
 394         if(jidx<j_index_end)
 395         {
 396
 397             jnrA             = jjnr[jidx];
 398             j_coord_offsetA  = DIM*jnrA;
 399
 400             /* load j atom coordinates */
 401             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 402                                               &jx0,&jy0,&jz0);
 403
 404             /* Calculate displacement vector */
 405             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 406             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 407             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 408             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 409             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 410             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 411             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 412             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 413             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 414
 415             /* Calculate squared distance and things based on it */
 416             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 417             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 418             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 419
 420             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 421             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 422             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 423
 424             /* Load parameters for j particles */
 425             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 426             vdwjidx0A        = 2*vdwtype[jnrA+0];
 427
 428             fjx0             = _fjsp_setzero_v2r8();
 429             fjy0             = _fjsp_setzero_v2r8();
 430             fjz0             = _fjsp_setzero_v2r8();
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 437
 438             /* Compute parameters for interactions between i and j atoms */
 439             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 440             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 441                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 442
 443             /* Calculate table index by multiplying r with table scale and truncate to integer */
 444             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 445             itab_tmp         = _fjsp_dtox_v2r8(rt);
 446             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 447             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 448             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 449
 450             vfconv.i[0]     *= 12;
 451             vfconv.i[1]     *= 12;
 452
 453             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 454             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 455             F                = _fjsp_setzero_v2r8();
 456             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 457             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 458             H                = _fjsp_setzero_v2r8();
 459             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 460             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 461             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 462             velec            = _fjsp_mul_v2r8(qq00,VV);
 463             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 464             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 465
 466             /* CUBIC SPLINE TABLE DISPERSION */
 467             vfconv.i[0]       += 4;
 468             vfconv.i[1]       += 4;
 469             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 470             F                = _fjsp_setzero_v2r8();
 471             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 472             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 473             H                = _fjsp_setzero_v2r8();
 474             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 475             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 476             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 477             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 478             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 479             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 480
 481             /* CUBIC SPLINE TABLE REPULSION */
 482             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 483             F                = _fjsp_setzero_v2r8();
 484             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 485             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 486             H                = _fjsp_setzero_v2r8();
 487             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 488             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 489             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 490             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 491             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 492             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 493             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 494             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 495
 496             /* Update potential sum for this i atom from the interaction with this j atom. */
 497             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 498             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 499             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 500             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 501
 502             fscal            = _fjsp_add_v2r8(felec,fvdw);
 503
 504             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 505
 506             /* Update vectorial force */
 507             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 508             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 509             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 510
 511             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 512             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 513             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 514
 515             /**************************
 516              * CALCULATE INTERACTIONS *
 517              **************************/
 518
 519             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 520
 521             /* Compute parameters for interactions between i and j atoms */
 522             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 523
 524             /* Calculate table index by multiplying r with table scale and truncate to integer */
 525             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 526             itab_tmp         = _fjsp_dtox_v2r8(rt);
 527             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 528             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 529             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 530
 531             vfconv.i[0]     *= 12;
 532             vfconv.i[1]     *= 12;
 533
 534             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 535             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 536             F                = _fjsp_setzero_v2r8();
 537             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 538             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 539             H                = _fjsp_setzero_v2r8();
 540             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 541             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 542             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 543             velec            = _fjsp_mul_v2r8(qq10,VV);
 544             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 545             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 546
 547             /* Update potential sum for this i atom from the interaction with this j atom. */
 548             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 549             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 550
 551             fscal            = felec;
 552
 553             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 554
 555             /* Update vectorial force */
 556             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 557             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 558             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 559
 560             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 561             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 562             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 563
 564             /**************************
 565              * CALCULATE INTERACTIONS *
 566              **************************/
 567
 568             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 569
 570             /* Compute parameters for interactions between i and j atoms */
 571             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 572
 573             /* Calculate table index by multiplying r with table scale and truncate to integer */
 574             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 575             itab_tmp         = _fjsp_dtox_v2r8(rt);
 576             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 577             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 578             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 579
 580             vfconv.i[0]     *= 12;
 581             vfconv.i[1]     *= 12;
 582
 583             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 584             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 585             F                = _fjsp_setzero_v2r8();
 586             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 587             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 588             H                = _fjsp_setzero_v2r8();
 589             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 590             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 591             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 592             velec            = _fjsp_mul_v2r8(qq20,VV);
 593             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 594             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 595
 596             /* Update potential sum for this i atom from the interaction with this j atom. */
 597             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 598             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 599
 600             fscal            = felec;
 601
 602             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 603
 604             /* Update vectorial force */
 605             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 606             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 607             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 608
 609             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 610             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 611             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 612
 613             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 614
 615             /* Inner loop uses 171 flops */
 616         }
 617
 618         /* End of innermost loop */
 619
 620         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 621                                               f+i_coord_offset,fshift+i_shift_offset);
 622
 623         ggid                        = gid[iidx];
 624         /* Update potential energies */
 625         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 626         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 627
 628         /* Increment number of inner iterations */
 629         inneriter                  += j_index_end - j_index_start;
 630
 631         /* Outer loop uses 20 flops */
 632     }
 633
 634     /* Increment number of outer iterations */
 635     outeriter        += nri;
 636
 637     /* Update outer/inner flops */
 638
 639     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*171);
 640 }
 641 /*
 642  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 643  * Electrostatics interaction: CubicSplineTable
 644  * VdW interaction:            CubicSplineTable
 645  * Geometry:                   Water3-Particle
 646  * Calculate force/pot:        Force
 647  */
 648 void
 649 nb_kernel_ElecCSTab_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 650                     (t_nblist                    * gmx_restrict       nlist,
 651                      rvec                        * gmx_restrict          xx,
 652                      rvec                        * gmx_restrict          ff,
 653                      t_forcerec                  * gmx_restrict          fr,
 654                      t_mdatoms                   * gmx_restrict     mdatoms,
 655                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 656                      t_nrnb                      * gmx_restrict        nrnb)
 657 {
 658     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 659      * just 0 for non-waters.
 660      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 661      * jnr indices corresponding to data put in the four positions in the SIMD register.
 662      */
 663     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 664     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 665     int              jnrA,jnrB;
 666     int              j_coord_offsetA,j_coord_offsetB;
 667     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 668     real             rcutoff_scalar;
 669     real             *shiftvec,*fshift,*x,*f;
 670     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 671     int              vdwioffset0;
 672     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 673     int              vdwioffset1;
 674     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 675     int              vdwioffset2;
 676     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 677     int              vdwjidx0A,vdwjidx0B;
 678     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 679     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 680     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 681     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 682     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 683     real             *charge;
 684     int              nvdwtype;
 685     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 686     int              *vdwtype;
 687     real             *vdwparam;
 688     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 689     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 690     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 691     real             *vftab;
 692     _fjsp_v2r8       itab_tmp;
 693     _fjsp_v2r8       dummy_mask,cutoff_mask;
 694     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 695     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 696     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 697
 698     x                = xx[0];
 699     f                = ff[0];
 700
 701     nri              = nlist->nri;
 702     iinr             = nlist->iinr;
 703     jindex           = nlist->jindex;
 704     jjnr             = nlist->jjnr;
 705     shiftidx         = nlist->shift;
 706     gid              = nlist->gid;
 707     shiftvec         = fr->shift_vec[0];
 708     fshift           = fr->fshift[0];
 709     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 710     charge           = mdatoms->chargeA;
 711     nvdwtype         = fr->ntype;
 712     vdwparam         = fr->nbfp;
 713     vdwtype          = mdatoms->typeA;
 714
 715     vftab            = kernel_data->table_elec_vdw->data;
 716     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec_vdw->scale);
 717
 718     /* Setup water-specific parameters */
 719     inr              = nlist->iinr[0];
 720     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 721     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 722     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 723     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 724
 725     /* Avoid stupid compiler warnings */
 726     jnrA = jnrB = 0;
 727     j_coord_offsetA = 0;
 728     j_coord_offsetB = 0;
 729
 730     outeriter        = 0;
 731     inneriter        = 0;
 732
 733     /* Start outer loop over neighborlists */
 734     for(iidx=0; iidx<nri; iidx++)
 735     {
 736         /* Load shift vector for this list */
 737         i_shift_offset   = DIM*shiftidx[iidx];
 738
 739         /* Load limits for loop over neighbors */
 740         j_index_start    = jindex[iidx];
 741         j_index_end      = jindex[iidx+1];
 742
 743         /* Get outer coordinate index */
 744         inr              = iinr[iidx];
 745         i_coord_offset   = DIM*inr;
 746
 747         /* Load i particle coords and add shift vector */
 748         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 749                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 750
 751         fix0             = _fjsp_setzero_v2r8();
 752         fiy0             = _fjsp_setzero_v2r8();
 753         fiz0             = _fjsp_setzero_v2r8();
 754         fix1             = _fjsp_setzero_v2r8();
 755         fiy1             = _fjsp_setzero_v2r8();
 756         fiz1             = _fjsp_setzero_v2r8();
 757         fix2             = _fjsp_setzero_v2r8();
 758         fiy2             = _fjsp_setzero_v2r8();
 759         fiz2             = _fjsp_setzero_v2r8();
 760
 761         /* Start inner kernel loop */
 762         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 763         {
 764
 765             /* Get j neighbor index, and coordinate index */
 766             jnrA             = jjnr[jidx];
 767             jnrB             = jjnr[jidx+1];
 768             j_coord_offsetA  = DIM*jnrA;
 769             j_coord_offsetB  = DIM*jnrB;
 770
 771             /* load j atom coordinates */
 772             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 773                                               &jx0,&jy0,&jz0);
 774
 775             /* Calculate displacement vector */
 776             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 777             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 778             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 779             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 780             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 781             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 782             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 783             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 784             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 785
 786             /* Calculate squared distance and things based on it */
 787             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 788             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 789             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 790
 791             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 792             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 793             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 794
 795             /* Load parameters for j particles */
 796             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 797             vdwjidx0A        = 2*vdwtype[jnrA+0];
 798             vdwjidx0B        = 2*vdwtype[jnrB+0];
 799
 800             fjx0             = _fjsp_setzero_v2r8();
 801             fjy0             = _fjsp_setzero_v2r8();
 802             fjz0             = _fjsp_setzero_v2r8();
 803
 804             /**************************
 805              * CALCULATE INTERACTIONS *
 806              **************************/
 807
 808             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 809
 810             /* Compute parameters for interactions between i and j atoms */
 811             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 812             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 813                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 814
 815             /* Calculate table index by multiplying r with table scale and truncate to integer */
 816             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 817             itab_tmp         = _fjsp_dtox_v2r8(rt);
 818             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 819             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 820             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 821
 822             vfconv.i[0]     *= 12;
 823             vfconv.i[1]     *= 12;
 824
 825             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 826             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 827             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 828             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 829             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 830             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 831             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 832             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 833             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 834             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
 835
 836             /* CUBIC SPLINE TABLE DISPERSION */
 837             vfconv.i[0]       += 4;
 838             vfconv.i[1]       += 4;
 839             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 840             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 841             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 842             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 843             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 844             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 845             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 846             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 847             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 848
 849             /* CUBIC SPLINE TABLE REPULSION */
 850             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 851             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 852             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 853             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 854             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 855             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 856             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 857             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 858             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 859             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 860
 861             fscal            = _fjsp_add_v2r8(felec,fvdw);
 862
 863             /* Update vectorial force */
 864             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 865             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 866             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 867
 868             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 869             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 870             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 871
 872             /**************************
 873              * CALCULATE INTERACTIONS *
 874              **************************/
 875
 876             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 877
 878             /* Compute parameters for interactions between i and j atoms */
 879             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 880
 881             /* Calculate table index by multiplying r with table scale and truncate to integer */
 882             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 883             itab_tmp         = _fjsp_dtox_v2r8(rt);
 884             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 885             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 886             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 887
 888             vfconv.i[0]     *= 12;
 889             vfconv.i[1]     *= 12;
 890
 891             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 892             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 893             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 894             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 895             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 896             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 897             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 898             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 899             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 900             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 901
 902             fscal            = felec;
 903
 904             /* Update vectorial force */
 905             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 906             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 907             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 908
 909             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 910             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 911             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 912
 913             /**************************
 914              * CALCULATE INTERACTIONS *
 915              **************************/
 916
 917             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 918
 919             /* Compute parameters for interactions between i and j atoms */
 920             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 921
 922             /* Calculate table index by multiplying r with table scale and truncate to integer */
 923             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 924             itab_tmp         = _fjsp_dtox_v2r8(rt);
 925             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 926             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 927             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 928
 929             vfconv.i[0]     *= 12;
 930             vfconv.i[1]     *= 12;
 931
 932             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 933             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 934             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 935             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 936             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 937             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 938             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 939             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 940             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 941             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 942
 943             fscal            = felec;
 944
 945             /* Update vectorial force */
 946             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 947             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 948             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 949
 950             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 951             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 952             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 953
 954             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 955
 956             /* Inner loop uses 151 flops */
 957         }
 958
 959         if(jidx<j_index_end)
 960         {
 961
 962             jnrA             = jjnr[jidx];
 963             j_coord_offsetA  = DIM*jnrA;
 964
 965             /* load j atom coordinates */
 966             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 967                                               &jx0,&jy0,&jz0);
 968
 969             /* Calculate displacement vector */
 970             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 971             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 972             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 973             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 974             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 975             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 976             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 977             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 978             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 979
 980             /* Calculate squared distance and things based on it */
 981             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 982             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 983             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 984
 985             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 986             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 987             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 988
 989             /* Load parameters for j particles */
 990             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 991             vdwjidx0A        = 2*vdwtype[jnrA+0];
 992
 993             fjx0             = _fjsp_setzero_v2r8();
 994             fjy0             = _fjsp_setzero_v2r8();
 995             fjz0             = _fjsp_setzero_v2r8();
 996
 997             /**************************
 998              * CALCULATE INTERACTIONS *
 999              **************************/
1000
1001             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1002
1003             /* Compute parameters for interactions between i and j atoms */
1004             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1005             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1006                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1007
1008             /* Calculate table index by multiplying r with table scale and truncate to integer */
1009             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1010             itab_tmp         = _fjsp_dtox_v2r8(rt);
1011             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1012             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1013             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1014
1015             vfconv.i[0]     *= 12;
1016             vfconv.i[1]     *= 12;
1017
1018             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1019             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1020             F                = _fjsp_setzero_v2r8();
1021             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1022             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1023             H                = _fjsp_setzero_v2r8();
1024             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1025             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1026             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1027             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,FF),_fjsp_mul_v2r8(vftabscale,rinv00)));
1028
1029             /* CUBIC SPLINE TABLE DISPERSION */
1030             vfconv.i[0]       += 4;
1031             vfconv.i[1]       += 4;
1032             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1033             F                = _fjsp_setzero_v2r8();
1034             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1035             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1036             H                = _fjsp_setzero_v2r8();
1037             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1038             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1039             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1040             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1041
1042             /* CUBIC SPLINE TABLE REPULSION */
1043             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1044             F                = _fjsp_setzero_v2r8();
1045             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1046             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1047             H                = _fjsp_setzero_v2r8();
1048             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1049             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1050             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1051             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1052             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1053
1054             fscal            = _fjsp_add_v2r8(felec,fvdw);
1055
1056             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1057
1058             /* Update vectorial force */
1059             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1060             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1061             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1062
1063             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1064             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1065             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1066
1067             /**************************
1068              * CALCULATE INTERACTIONS *
1069              **************************/
1070
1071             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1072
1073             /* Compute parameters for interactions between i and j atoms */
1074             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1075
1076             /* Calculate table index by multiplying r with table scale and truncate to integer */
1077             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1078             itab_tmp         = _fjsp_dtox_v2r8(rt);
1079             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1080             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1081             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1082
1083             vfconv.i[0]     *= 12;
1084             vfconv.i[1]     *= 12;
1085
1086             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1087             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1088             F                = _fjsp_setzero_v2r8();
1089             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1090             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1091             H                = _fjsp_setzero_v2r8();
1092             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1093             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1094             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1095             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1096
1097             fscal            = felec;
1098
1099             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1100
1101             /* Update vectorial force */
1102             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1103             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1104             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1105
1106             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1107             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1108             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1109
1110             /**************************
1111              * CALCULATE INTERACTIONS *
1112              **************************/
1113
1114             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1115
1116             /* Compute parameters for interactions between i and j atoms */
1117             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1118
1119             /* Calculate table index by multiplying r with table scale and truncate to integer */
1120             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1121             itab_tmp         = _fjsp_dtox_v2r8(rt);
1122             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1123             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1124             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1125
1126             vfconv.i[0]     *= 12;
1127             vfconv.i[1]     *= 12;
1128
1129             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1130             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1131             F                = _fjsp_setzero_v2r8();
1132             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1133             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1134             H                = _fjsp_setzero_v2r8();
1135             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1136             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1137             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1138             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1139
1140             fscal            = felec;
1141
1142             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1143
1144             /* Update vectorial force */
1145             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1146             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1147             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1148
1149             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1150             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1151             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1152
1153             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1154
1155             /* Inner loop uses 151 flops */
1156         }
1157
1158         /* End of innermost loop */
1159
1160         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1161                                               f+i_coord_offset,fshift+i_shift_offset);
1162
1163         /* Increment number of inner iterations */
1164         inneriter                  += j_index_end - j_index_start;
1165
1166         /* Outer loop uses 18 flops */
1167     }
1168
1169     /* Increment number of outer iterations */
1170     outeriter        += nri;
1171
1172     /* Update outer/inner flops */
1173
1174     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*151);
1175 }