src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: CubicSplineTable
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Water4-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_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              vdwioffset3;
  88     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  89     int              vdwjidx0A,vdwjidx0B;
  90     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  91     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  92     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  93     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  94     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  95     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  96     real             *charge;
  97     int              nvdwtype;
  98     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  99     int              *vdwtype;
 100     real             *vdwparam;
 101     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 102     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 103     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 104     real             *vftab;
 105     _fjsp_v2r8       itab_tmp;
 106     _fjsp_v2r8       dummy_mask,cutoff_mask;
 107     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 108     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 109     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 110
 111     x                = xx[0];
 112     f                = ff[0];
 113
 114     nri              = nlist->nri;
 115     iinr             = nlist->iinr;
 116     jindex           = nlist->jindex;
 117     jjnr             = nlist->jjnr;
 118     shiftidx         = nlist->shift;
 119     gid              = nlist->gid;
 120     shiftvec         = fr->shift_vec[0];
 121     fshift           = fr->fshift[0];
 122     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 123     charge           = mdatoms->chargeA;
 124     nvdwtype         = fr->ntype;
 125     vdwparam         = fr->nbfp;
 126     vdwtype          = mdatoms->typeA;
 127
 128     vftab            = kernel_data->table_elec->data;
 129     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 134     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 135     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     /* Avoid stupid compiler warnings */
 139     jnrA = jnrB = 0;
 140     j_coord_offsetA = 0;
 141     j_coord_offsetB = 0;
 142
 143     outeriter        = 0;
 144     inneriter        = 0;
 145
 146     /* Start outer loop over neighborlists */
 147     for(iidx=0; iidx<nri; iidx++)
 148     {
 149         /* Load shift vector for this list */
 150         i_shift_offset   = DIM*shiftidx[iidx];
 151
 152         /* Load limits for loop over neighbors */
 153         j_index_start    = jindex[iidx];
 154         j_index_end      = jindex[iidx+1];
 155
 156         /* Get outer coordinate index */
 157         inr              = iinr[iidx];
 158         i_coord_offset   = DIM*inr;
 159
 160         /* Load i particle coords and add shift vector */
 161         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 162                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 163
 164         fix0             = _fjsp_setzero_v2r8();
 165         fiy0             = _fjsp_setzero_v2r8();
 166         fiz0             = _fjsp_setzero_v2r8();
 167         fix1             = _fjsp_setzero_v2r8();
 168         fiy1             = _fjsp_setzero_v2r8();
 169         fiz1             = _fjsp_setzero_v2r8();
 170         fix2             = _fjsp_setzero_v2r8();
 171         fiy2             = _fjsp_setzero_v2r8();
 172         fiz2             = _fjsp_setzero_v2r8();
 173         fix3             = _fjsp_setzero_v2r8();
 174         fiy3             = _fjsp_setzero_v2r8();
 175         fiz3             = _fjsp_setzero_v2r8();
 176
 177         /* Reset potential sums */
 178         velecsum         = _fjsp_setzero_v2r8();
 179         vvdwsum          = _fjsp_setzero_v2r8();
 180
 181         /* Start inner kernel loop */
 182         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 183         {
 184
 185             /* Get j neighbor index, and coordinate index */
 186             jnrA             = jjnr[jidx];
 187             jnrB             = jjnr[jidx+1];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190
 191             /* load j atom coordinates */
 192             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 193                                               &jx0,&jy0,&jz0);
 194
 195             /* Calculate displacement vector */
 196             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 197             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 198             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 199             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 200             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 201             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 202             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 203             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 204             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 205             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 206             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 207             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 208
 209             /* Calculate squared distance and things based on it */
 210             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 211             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 212             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 213             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 214
 215             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 216             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 217             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 218
 219             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 220
 221             /* Load parameters for j particles */
 222             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 223             vdwjidx0A        = 2*vdwtype[jnrA+0];
 224             vdwjidx0B        = 2*vdwtype[jnrB+0];
 225
 226             fjx0             = _fjsp_setzero_v2r8();
 227             fjy0             = _fjsp_setzero_v2r8();
 228             fjz0             = _fjsp_setzero_v2r8();
 229
 230             /**************************
 231              * CALCULATE INTERACTIONS *
 232              **************************/
 233
 234             /* Compute parameters for interactions between i and j atoms */
 235             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 236                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 237
 238             /* LENNARD-JONES DISPERSION/REPULSION */
 239
 240             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 241             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 242             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 243             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 244             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 245
 246             /* Update potential sum for this i atom from the interaction with this j atom. */
 247             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 248
 249             fscal            = fvdw;
 250
 251             /* Update vectorial force */
 252             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 253             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 254             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 255
 256             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 257             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 258             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 259
 260             /**************************
 261              * CALCULATE INTERACTIONS *
 262              **************************/
 263
 264             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 265
 266             /* Compute parameters for interactions between i and j atoms */
 267             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 268
 269             /* Calculate table index by multiplying r with table scale and truncate to integer */
 270             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 271             itab_tmp         = _fjsp_dtox_v2r8(rt);
 272             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 273             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 274             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 275
 276             vfconv.i[0]     *= 4;
 277             vfconv.i[1]     *= 4;
 278
 279             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 280             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 281             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 282             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 283             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 284             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 285             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 286             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 287             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 288             velec            = _fjsp_mul_v2r8(qq10,VV);
 289             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 290             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 291
 292             /* Update potential sum for this i atom from the interaction with this j atom. */
 293             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 294
 295             fscal            = felec;
 296
 297             /* Update vectorial force */
 298             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 299             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 300             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 301
 302             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 303             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 304             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 305
 306             /**************************
 307              * CALCULATE INTERACTIONS *
 308              **************************/
 309
 310             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 311
 312             /* Compute parameters for interactions between i and j atoms */
 313             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 314
 315             /* Calculate table index by multiplying r with table scale and truncate to integer */
 316             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 317             itab_tmp         = _fjsp_dtox_v2r8(rt);
 318             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 319             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 320             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 321
 322             vfconv.i[0]     *= 4;
 323             vfconv.i[1]     *= 4;
 324
 325             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 326             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 327             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 328             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 329             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 330             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 331             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 332             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 333             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 334             velec            = _fjsp_mul_v2r8(qq20,VV);
 335             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 336             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 337
 338             /* Update potential sum for this i atom from the interaction with this j atom. */
 339             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 340
 341             fscal            = felec;
 342
 343             /* Update vectorial force */
 344             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 345             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 346             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 347
 348             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 349             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 350             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 351
 352             /**************************
 353              * CALCULATE INTERACTIONS *
 354              **************************/
 355
 356             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 357
 358             /* Compute parameters for interactions between i and j atoms */
 359             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 360
 361             /* Calculate table index by multiplying r with table scale and truncate to integer */
 362             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 363             itab_tmp         = _fjsp_dtox_v2r8(rt);
 364             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 365             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 366             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 367
 368             vfconv.i[0]     *= 4;
 369             vfconv.i[1]     *= 4;
 370
 371             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 372             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 373             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 374             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 375             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 376             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 377             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 378             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 379             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 380             velec            = _fjsp_mul_v2r8(qq30,VV);
 381             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 382             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 383
 384             /* Update potential sum for this i atom from the interaction with this j atom. */
 385             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 386
 387             fscal            = felec;
 388
 389             /* Update vectorial force */
 390             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 391             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 392             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 393
 394             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 395             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 396             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 397
 398             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 399
 400             /* Inner loop uses 176 flops */
 401         }
 402
 403         if(jidx<j_index_end)
 404         {
 405
 406             jnrA             = jjnr[jidx];
 407             j_coord_offsetA  = DIM*jnrA;
 408
 409             /* load j atom coordinates */
 410             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 411                                               &jx0,&jy0,&jz0);
 412
 413             /* Calculate displacement vector */
 414             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 415             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 416             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 417             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 418             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 419             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 420             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 421             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 422             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 423             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 424             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 425             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 426
 427             /* Calculate squared distance and things based on it */
 428             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 429             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 430             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 431             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 432
 433             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 434             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 435             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 436
 437             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 438
 439             /* Load parameters for j particles */
 440             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 441             vdwjidx0A        = 2*vdwtype[jnrA+0];
 442
 443             fjx0             = _fjsp_setzero_v2r8();
 444             fjy0             = _fjsp_setzero_v2r8();
 445             fjz0             = _fjsp_setzero_v2r8();
 446
 447             /**************************
 448              * CALCULATE INTERACTIONS *
 449              **************************/
 450
 451             /* Compute parameters for interactions between i and j atoms */
 452             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 453                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 454
 455             /* LENNARD-JONES DISPERSION/REPULSION */
 456
 457             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 458             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 459             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 460             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 461             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 462
 463             /* Update potential sum for this i atom from the interaction with this j atom. */
 464             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 465             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 466
 467             fscal            = fvdw;
 468
 469             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 470
 471             /* Update vectorial force */
 472             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 473             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 474             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 475
 476             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 477             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 478             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 479
 480             /**************************
 481              * CALCULATE INTERACTIONS *
 482              **************************/
 483
 484             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 485
 486             /* Compute parameters for interactions between i and j atoms */
 487             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 488
 489             /* Calculate table index by multiplying r with table scale and truncate to integer */
 490             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 491             itab_tmp         = _fjsp_dtox_v2r8(rt);
 492             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 493             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 494             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 495
 496             vfconv.i[0]     *= 4;
 497             vfconv.i[1]     *= 4;
 498
 499             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 500             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 501             F                = _fjsp_setzero_v2r8();
 502             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 503             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 504             H                = _fjsp_setzero_v2r8();
 505             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 506             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 507             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 508             velec            = _fjsp_mul_v2r8(qq10,VV);
 509             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 510             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 511
 512             /* Update potential sum for this i atom from the interaction with this j atom. */
 513             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 514             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 515
 516             fscal            = felec;
 517
 518             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 519
 520             /* Update vectorial force */
 521             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 522             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 523             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 524
 525             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 526             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 527             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 528
 529             /**************************
 530              * CALCULATE INTERACTIONS *
 531              **************************/
 532
 533             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 534
 535             /* Compute parameters for interactions between i and j atoms */
 536             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 537
 538             /* Calculate table index by multiplying r with table scale and truncate to integer */
 539             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 540             itab_tmp         = _fjsp_dtox_v2r8(rt);
 541             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 542             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 543             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 544
 545             vfconv.i[0]     *= 4;
 546             vfconv.i[1]     *= 4;
 547
 548             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 549             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 550             F                = _fjsp_setzero_v2r8();
 551             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 552             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 553             H                = _fjsp_setzero_v2r8();
 554             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 555             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 556             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 557             velec            = _fjsp_mul_v2r8(qq20,VV);
 558             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 559             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 560
 561             /* Update potential sum for this i atom from the interaction with this j atom. */
 562             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 563             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 564
 565             fscal            = felec;
 566
 567             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 568
 569             /* Update vectorial force */
 570             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 571             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 572             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 573
 574             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 575             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 576             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 577
 578             /**************************
 579              * CALCULATE INTERACTIONS *
 580              **************************/
 581
 582             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 583
 584             /* Compute parameters for interactions between i and j atoms */
 585             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 586
 587             /* Calculate table index by multiplying r with table scale and truncate to integer */
 588             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 589             itab_tmp         = _fjsp_dtox_v2r8(rt);
 590             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 591             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 592             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 593
 594             vfconv.i[0]     *= 4;
 595             vfconv.i[1]     *= 4;
 596
 597             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 598             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 599             F                = _fjsp_setzero_v2r8();
 600             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 601             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 602             H                = _fjsp_setzero_v2r8();
 603             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 604             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 605             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 606             velec            = _fjsp_mul_v2r8(qq30,VV);
 607             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 608             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 609
 610             /* Update potential sum for this i atom from the interaction with this j atom. */
 611             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 612             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 613
 614             fscal            = felec;
 615
 616             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 617
 618             /* Update vectorial force */
 619             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 620             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 621             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 622
 623             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 624             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 625             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 626
 627             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 628
 629             /* Inner loop uses 176 flops */
 630         }
 631
 632         /* End of innermost loop */
 633
 634         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 635                                               f+i_coord_offset,fshift+i_shift_offset);
 636
 637         ggid                        = gid[iidx];
 638         /* Update potential energies */
 639         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 640         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 641
 642         /* Increment number of inner iterations */
 643         inneriter                  += j_index_end - j_index_start;
 644
 645         /* Outer loop uses 26 flops */
 646     }
 647
 648     /* Increment number of outer iterations */
 649     outeriter        += nri;
 650
 651     /* Update outer/inner flops */
 652
 653     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*176);
 654 }
 655 /*
 656  * Gromacs nonbonded kernel:   nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 657  * Electrostatics interaction: CubicSplineTable
 658  * VdW interaction:            LennardJones
 659  * Geometry:                   Water4-Particle
 660  * Calculate force/pot:        Force
 661  */
 662 void
 663 nb_kernel_ElecCSTab_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 664                     (t_nblist                    * gmx_restrict       nlist,
 665                      rvec                        * gmx_restrict          xx,
 666                      rvec                        * gmx_restrict          ff,
 667                      t_forcerec                  * gmx_restrict          fr,
 668                      t_mdatoms                   * gmx_restrict     mdatoms,
 669                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 670                      t_nrnb                      * gmx_restrict        nrnb)
 671 {
 672     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 673      * just 0 for non-waters.
 674      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 675      * jnr indices corresponding to data put in the four positions in the SIMD register.
 676      */
 677     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 678     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 679     int              jnrA,jnrB;
 680     int              j_coord_offsetA,j_coord_offsetB;
 681     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 682     real             rcutoff_scalar;
 683     real             *shiftvec,*fshift,*x,*f;
 684     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 685     int              vdwioffset0;
 686     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 687     int              vdwioffset1;
 688     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 689     int              vdwioffset2;
 690     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 691     int              vdwioffset3;
 692     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 693     int              vdwjidx0A,vdwjidx0B;
 694     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 695     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 696     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 697     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 698     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 699     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 700     real             *charge;
 701     int              nvdwtype;
 702     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 703     int              *vdwtype;
 704     real             *vdwparam;
 705     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 706     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 707     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 708     real             *vftab;
 709     _fjsp_v2r8       itab_tmp;
 710     _fjsp_v2r8       dummy_mask,cutoff_mask;
 711     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 712     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 713     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 714
 715     x                = xx[0];
 716     f                = ff[0];
 717
 718     nri              = nlist->nri;
 719     iinr             = nlist->iinr;
 720     jindex           = nlist->jindex;
 721     jjnr             = nlist->jjnr;
 722     shiftidx         = nlist->shift;
 723     gid              = nlist->gid;
 724     shiftvec         = fr->shift_vec[0];
 725     fshift           = fr->fshift[0];
 726     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 727     charge           = mdatoms->chargeA;
 728     nvdwtype         = fr->ntype;
 729     vdwparam         = fr->nbfp;
 730     vdwtype          = mdatoms->typeA;
 731
 732     vftab            = kernel_data->table_elec->data;
 733     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_elec->scale);
 734
 735     /* Setup water-specific parameters */
 736     inr              = nlist->iinr[0];
 737     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 738     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 739     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 740     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 741
 742     /* Avoid stupid compiler warnings */
 743     jnrA = jnrB = 0;
 744     j_coord_offsetA = 0;
 745     j_coord_offsetB = 0;
 746
 747     outeriter        = 0;
 748     inneriter        = 0;
 749
 750     /* Start outer loop over neighborlists */
 751     for(iidx=0; iidx<nri; iidx++)
 752     {
 753         /* Load shift vector for this list */
 754         i_shift_offset   = DIM*shiftidx[iidx];
 755
 756         /* Load limits for loop over neighbors */
 757         j_index_start    = jindex[iidx];
 758         j_index_end      = jindex[iidx+1];
 759
 760         /* Get outer coordinate index */
 761         inr              = iinr[iidx];
 762         i_coord_offset   = DIM*inr;
 763
 764         /* Load i particle coords and add shift vector */
 765         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 766                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 767
 768         fix0             = _fjsp_setzero_v2r8();
 769         fiy0             = _fjsp_setzero_v2r8();
 770         fiz0             = _fjsp_setzero_v2r8();
 771         fix1             = _fjsp_setzero_v2r8();
 772         fiy1             = _fjsp_setzero_v2r8();
 773         fiz1             = _fjsp_setzero_v2r8();
 774         fix2             = _fjsp_setzero_v2r8();
 775         fiy2             = _fjsp_setzero_v2r8();
 776         fiz2             = _fjsp_setzero_v2r8();
 777         fix3             = _fjsp_setzero_v2r8();
 778         fiy3             = _fjsp_setzero_v2r8();
 779         fiz3             = _fjsp_setzero_v2r8();
 780
 781         /* Start inner kernel loop */
 782         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 783         {
 784
 785             /* Get j neighbor index, and coordinate index */
 786             jnrA             = jjnr[jidx];
 787             jnrB             = jjnr[jidx+1];
 788             j_coord_offsetA  = DIM*jnrA;
 789             j_coord_offsetB  = DIM*jnrB;
 790
 791             /* load j atom coordinates */
 792             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 793                                               &jx0,&jy0,&jz0);
 794
 795             /* Calculate displacement vector */
 796             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 797             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 798             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 799             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 800             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 801             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 802             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 803             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 804             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 805             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 806             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 807             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 808
 809             /* Calculate squared distance and things based on it */
 810             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 811             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 812             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 813             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 814
 815             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 816             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 817             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 818
 819             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 820
 821             /* Load parameters for j particles */
 822             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 823             vdwjidx0A        = 2*vdwtype[jnrA+0];
 824             vdwjidx0B        = 2*vdwtype[jnrB+0];
 825
 826             fjx0             = _fjsp_setzero_v2r8();
 827             fjy0             = _fjsp_setzero_v2r8();
 828             fjz0             = _fjsp_setzero_v2r8();
 829
 830             /**************************
 831              * CALCULATE INTERACTIONS *
 832              **************************/
 833
 834             /* Compute parameters for interactions between i and j atoms */
 835             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 836                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 837
 838             /* LENNARD-JONES DISPERSION/REPULSION */
 839
 840             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 841             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 842
 843             fscal            = fvdw;
 844
 845             /* Update vectorial force */
 846             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 847             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 848             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 849
 850             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 851             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 852             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 853
 854             /**************************
 855              * CALCULATE INTERACTIONS *
 856              **************************/
 857
 858             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 859
 860             /* Compute parameters for interactions between i and j atoms */
 861             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 862
 863             /* Calculate table index by multiplying r with table scale and truncate to integer */
 864             rt               = _fjsp_mul_v2r8(r10,vftabscale);
 865             itab_tmp         = _fjsp_dtox_v2r8(rt);
 866             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 867             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 868             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 869
 870             vfconv.i[0]     *= 4;
 871             vfconv.i[1]     *= 4;
 872
 873             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 874             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 875             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 876             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 877             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 878             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 879             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 880             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 881             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 882             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
 883
 884             fscal            = felec;
 885
 886             /* Update vectorial force */
 887             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 888             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 889             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 890
 891             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 892             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 893             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 894
 895             /**************************
 896              * CALCULATE INTERACTIONS *
 897              **************************/
 898
 899             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 900
 901             /* Compute parameters for interactions between i and j atoms */
 902             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 903
 904             /* Calculate table index by multiplying r with table scale and truncate to integer */
 905             rt               = _fjsp_mul_v2r8(r20,vftabscale);
 906             itab_tmp         = _fjsp_dtox_v2r8(rt);
 907             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 908             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 909             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 910
 911             vfconv.i[0]     *= 4;
 912             vfconv.i[1]     *= 4;
 913
 914             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 915             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 916             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 917             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 918             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 919             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 920             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 921             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 922             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 923             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
 924
 925             fscal            = felec;
 926
 927             /* Update vectorial force */
 928             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 929             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 930             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 931
 932             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 933             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 934             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 935
 936             /**************************
 937              * CALCULATE INTERACTIONS *
 938              **************************/
 939
 940             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 941
 942             /* Compute parameters for interactions between i and j atoms */
 943             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 944
 945             /* Calculate table index by multiplying r with table scale and truncate to integer */
 946             rt               = _fjsp_mul_v2r8(r30,vftabscale);
 947             itab_tmp         = _fjsp_dtox_v2r8(rt);
 948             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 949             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 950             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 951
 952             vfconv.i[0]     *= 4;
 953             vfconv.i[1]     *= 4;
 954
 955             /* CUBIC SPLINE TABLE ELECTROSTATICS */
 956             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 957             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 958             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 959             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
 960             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] +2);
 961             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 962             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
 963             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
 964             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
 965
 966             fscal            = felec;
 967
 968             /* Update vectorial force */
 969             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 970             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 971             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 972
 973             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 974             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 975             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 976
 977             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 978
 979             /* Inner loop uses 159 flops */
 980         }
 981
 982         if(jidx<j_index_end)
 983         {
 984
 985             jnrA             = jjnr[jidx];
 986             j_coord_offsetA  = DIM*jnrA;
 987
 988             /* load j atom coordinates */
 989             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 990                                               &jx0,&jy0,&jz0);
 991
 992             /* Calculate displacement vector */
 993             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 994             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 995             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 996             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 997             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 998             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 999             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1000             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1001             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1002             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1003             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1004             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1005
1006             /* Calculate squared distance and things based on it */
1007             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1008             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1009             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1010             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1011
1012             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1013             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1014             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1015
1016             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
1017
1018             /* Load parameters for j particles */
1019             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1020             vdwjidx0A        = 2*vdwtype[jnrA+0];
1021
1022             fjx0             = _fjsp_setzero_v2r8();
1023             fjy0             = _fjsp_setzero_v2r8();
1024             fjz0             = _fjsp_setzero_v2r8();
1025
1026             /**************************
1027              * CALCULATE INTERACTIONS *
1028              **************************/
1029
1030             /* Compute parameters for interactions between i and j atoms */
1031             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1032                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1033
1034             /* LENNARD-JONES DISPERSION/REPULSION */
1035
1036             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1037             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1038
1039             fscal            = fvdw;
1040
1041             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1042
1043             /* Update vectorial force */
1044             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1045             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1046             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1047
1048             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1049             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1050             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1051
1052             /**************************
1053              * CALCULATE INTERACTIONS *
1054              **************************/
1055
1056             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1057
1058             /* Compute parameters for interactions between i and j atoms */
1059             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1060
1061             /* Calculate table index by multiplying r with table scale and truncate to integer */
1062             rt               = _fjsp_mul_v2r8(r10,vftabscale);
1063             itab_tmp         = _fjsp_dtox_v2r8(rt);
1064             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1065             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1066             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1067
1068             vfconv.i[0]     *= 4;
1069             vfconv.i[1]     *= 4;
1070
1071             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1072             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1073             F                = _fjsp_setzero_v2r8();
1074             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1075             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1076             H                = _fjsp_setzero_v2r8();
1077             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1078             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1079             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1080             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,FF),_fjsp_mul_v2r8(vftabscale,rinv10)));
1081
1082             fscal            = felec;
1083
1084             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1085
1086             /* Update vectorial force */
1087             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1088             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1089             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1090
1091             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1092             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1093             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1094
1095             /**************************
1096              * CALCULATE INTERACTIONS *
1097              **************************/
1098
1099             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1100
1101             /* Compute parameters for interactions between i and j atoms */
1102             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1103
1104             /* Calculate table index by multiplying r with table scale and truncate to integer */
1105             rt               = _fjsp_mul_v2r8(r20,vftabscale);
1106             itab_tmp         = _fjsp_dtox_v2r8(rt);
1107             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1108             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1109             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1110
1111             vfconv.i[0]     *= 4;
1112             vfconv.i[1]     *= 4;
1113
1114             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1115             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1116             F                = _fjsp_setzero_v2r8();
1117             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1118             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1119             H                = _fjsp_setzero_v2r8();
1120             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1121             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1122             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1123             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,FF),_fjsp_mul_v2r8(vftabscale,rinv20)));
1124
1125             fscal            = felec;
1126
1127             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1128
1129             /* Update vectorial force */
1130             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1131             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1132             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1133
1134             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1135             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1136             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1137
1138             /**************************
1139              * CALCULATE INTERACTIONS *
1140              **************************/
1141
1142             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1143
1144             /* Compute parameters for interactions between i and j atoms */
1145             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1146
1147             /* Calculate table index by multiplying r with table scale and truncate to integer */
1148             rt               = _fjsp_mul_v2r8(r30,vftabscale);
1149             itab_tmp         = _fjsp_dtox_v2r8(rt);
1150             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1151             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1152             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1153
1154             vfconv.i[0]     *= 4;
1155             vfconv.i[1]     *= 4;
1156
1157             /* CUBIC SPLINE TABLE ELECTROSTATICS */
1158             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1159             F                = _fjsp_setzero_v2r8();
1160             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1161             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] +2);
1162             H                = _fjsp_setzero_v2r8();
1163             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1164             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(vfeps,H,G),F);
1165             FF               = _fjsp_madd_v2r8(_fjsp_madd_v2r8(twovfeps,H,G),vfeps,Fp);
1166             felec            = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,FF),_fjsp_mul_v2r8(vftabscale,rinv30)));
1167
1168             fscal            = felec;
1169
1170             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1171
1172             /* Update vectorial force */
1173             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1174             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1175             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1176
1177             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1178             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1179             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1180
1181             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1182
1183             /* Inner loop uses 159 flops */
1184         }
1185
1186         /* End of innermost loop */
1187
1188         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1189                                               f+i_coord_offset,fshift+i_shift_offset);
1190
1191         /* Increment number of inner iterations */
1192         inneriter                  += j_index_end - j_index_start;
1193
1194         /* Outer loop uses 24 flops */
1195     }
1196
1197     /* Increment number of outer iterations */
1198     outeriter        += nri;
1199
1200     /* Update outer/inner flops */
1201
1202     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*159);
1203 }