src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Water3-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEw_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       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 103     real             *ewtab;
 104     _fjsp_v2r8       itab_tmp;
 105     _fjsp_v2r8       dummy_mask,cutoff_mask;
 106     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 107     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 108     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 109
 110     x                = xx[0];
 111     f                = ff[0];
 112
 113     nri              = nlist->nri;
 114     iinr             = nlist->iinr;
 115     jindex           = nlist->jindex;
 116     jjnr             = nlist->jjnr;
 117     shiftidx         = nlist->shift;
 118     gid              = nlist->gid;
 119     shiftvec         = fr->shift_vec[0];
 120     fshift           = fr->fshift[0];
 121     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 122     charge           = mdatoms->chargeA;
 123     nvdwtype         = fr->ntype;
 124     vdwparam         = fr->nbfp;
 125     vdwtype          = mdatoms->typeA;
 126
 127     vftab            = kernel_data->table_vdw->data;
 128     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 129
 130     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 131     ewtab            = fr->ic->tabq_coul_FDV0;
 132     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 133     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 134
 135     /* Setup water-specific parameters */
 136     inr              = nlist->iinr[0];
 137     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 138     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 139     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 140     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 141
 142     /* Avoid stupid compiler warnings */
 143     jnrA = jnrB = 0;
 144     j_coord_offsetA = 0;
 145     j_coord_offsetB = 0;
 146
 147     outeriter        = 0;
 148     inneriter        = 0;
 149
 150     /* Start outer loop over neighborlists */
 151     for(iidx=0; iidx<nri; iidx++)
 152     {
 153         /* Load shift vector for this list */
 154         i_shift_offset   = DIM*shiftidx[iidx];
 155
 156         /* Load limits for loop over neighbors */
 157         j_index_start    = jindex[iidx];
 158         j_index_end      = jindex[iidx+1];
 159
 160         /* Get outer coordinate index */
 161         inr              = iinr[iidx];
 162         i_coord_offset   = DIM*inr;
 163
 164         /* Load i particle coords and add shift vector */
 165         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 166                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 167
 168         fix0             = _fjsp_setzero_v2r8();
 169         fiy0             = _fjsp_setzero_v2r8();
 170         fiz0             = _fjsp_setzero_v2r8();
 171         fix1             = _fjsp_setzero_v2r8();
 172         fiy1             = _fjsp_setzero_v2r8();
 173         fiz1             = _fjsp_setzero_v2r8();
 174         fix2             = _fjsp_setzero_v2r8();
 175         fiy2             = _fjsp_setzero_v2r8();
 176         fiz2             = _fjsp_setzero_v2r8();
 177
 178         /* Reset potential sums */
 179         velecsum         = _fjsp_setzero_v2r8();
 180         vvdwsum          = _fjsp_setzero_v2r8();
 181
 182         /* Start inner kernel loop */
 183         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 184         {
 185
 186             /* Get j neighbor index, and coordinate index */
 187             jnrA             = jjnr[jidx];
 188             jnrB             = jjnr[jidx+1];
 189             j_coord_offsetA  = DIM*jnrA;
 190             j_coord_offsetB  = DIM*jnrB;
 191
 192             /* load j atom coordinates */
 193             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 194                                               &jx0,&jy0,&jz0);
 195
 196             /* Calculate displacement vector */
 197             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 198             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 199             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 200             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 201             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 202             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 203             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 204             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 205             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 206
 207             /* Calculate squared distance and things based on it */
 208             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 209             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 210             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 211
 212             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 213             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 214             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 215
 216             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 217             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 218             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 219
 220             /* Load parameters for j particles */
 221             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 222             vdwjidx0A        = 2*vdwtype[jnrA+0];
 223             vdwjidx0B        = 2*vdwtype[jnrB+0];
 224
 225             fjx0             = _fjsp_setzero_v2r8();
 226             fjy0             = _fjsp_setzero_v2r8();
 227             fjz0             = _fjsp_setzero_v2r8();
 228
 229             /**************************
 230              * CALCULATE INTERACTIONS *
 231              **************************/
 232
 233             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 234
 235             /* Compute parameters for interactions between i and j atoms */
 236             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 237             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 238                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 239
 240             /* Calculate table index by multiplying r with table scale and truncate to integer */
 241             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 242             itab_tmp         = _fjsp_dtox_v2r8(rt);
 243             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 244             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 245             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 246
 247             vfconv.i[0]     *= 8;
 248             vfconv.i[1]     *= 8;
 249
 250             /* EWALD ELECTROSTATICS */
 251
 252             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 253             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 254             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 255             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 256             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 257
 258             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 259             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 260             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 261             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 262             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 263             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 264             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 265             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 266             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 267             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 268
 269             /* CUBIC SPLINE TABLE DISPERSION */
 270             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 271             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 272             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 273             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 274             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 276             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 277             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 278             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 279             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 280             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 281
 282             /* CUBIC SPLINE TABLE REPULSION */
 283             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 284             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 285             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 286             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 287             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 288             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 289             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 290             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 291             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 292             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 293             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 294             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 295             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 296
 297             /* Update potential sum for this i atom from the interaction with this j atom. */
 298             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 299             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 300
 301             fscal            = _fjsp_add_v2r8(felec,fvdw);
 302
 303             /* Update vectorial force */
 304             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 305             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 306             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 307
 308             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 309             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 310             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 311
 312             /**************************
 313              * CALCULATE INTERACTIONS *
 314              **************************/
 315
 316             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 317
 318             /* Compute parameters for interactions between i and j atoms */
 319             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 320
 321             /* EWALD ELECTROSTATICS */
 322
 323             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 324             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 325             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 326             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 327             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 328
 329             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 330             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 331             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 332             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 333             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 334             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 335             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 336             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 337             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 338             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 339
 340             /* Update potential sum for this i atom from the interaction with this j atom. */
 341             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 342
 343             fscal            = felec;
 344
 345             /* Update vectorial force */
 346             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 347             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 348             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 349
 350             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 351             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 352             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 353
 354             /**************************
 355              * CALCULATE INTERACTIONS *
 356              **************************/
 357
 358             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 359
 360             /* Compute parameters for interactions between i and j atoms */
 361             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 362
 363             /* EWALD ELECTROSTATICS */
 364
 365             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 366             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 367             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 368             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 369             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 370
 371             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 372             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 373             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 374             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 375             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 376             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 377             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 378             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 379             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 380             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 381
 382             /* Update potential sum for this i atom from the interaction with this j atom. */
 383             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 384
 385             fscal            = felec;
 386
 387             /* Update vectorial force */
 388             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 389             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 390             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 391
 392             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 393             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 394             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 395
 396             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 397
 398             /* Inner loop uses 169 flops */
 399         }
 400
 401         if(jidx<j_index_end)
 402         {
 403
 404             jnrA             = jjnr[jidx];
 405             j_coord_offsetA  = DIM*jnrA;
 406
 407             /* load j atom coordinates */
 408             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 409                                               &jx0,&jy0,&jz0);
 410
 411             /* Calculate displacement vector */
 412             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 413             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 414             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 415             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 416             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 417             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 418             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 419             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 420             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 421
 422             /* Calculate squared distance and things based on it */
 423             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 424             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 425             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 426
 427             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 428             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 429             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 430
 431             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 432             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 433             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 434
 435             /* Load parameters for j particles */
 436             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 437             vdwjidx0A        = 2*vdwtype[jnrA+0];
 438
 439             fjx0             = _fjsp_setzero_v2r8();
 440             fjy0             = _fjsp_setzero_v2r8();
 441             fjz0             = _fjsp_setzero_v2r8();
 442
 443             /**************************
 444              * CALCULATE INTERACTIONS *
 445              **************************/
 446
 447             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 448
 449             /* Compute parameters for interactions between i and j atoms */
 450             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 451             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 452                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 453
 454             /* Calculate table index by multiplying r with table scale and truncate to integer */
 455             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 456             itab_tmp         = _fjsp_dtox_v2r8(rt);
 457             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 458             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 459             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 460
 461             vfconv.i[0]     *= 8;
 462             vfconv.i[1]     *= 8;
 463
 464             /* EWALD ELECTROSTATICS */
 465
 466             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 467             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 468             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 469             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 470             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 471
 472             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 473             ewtabD           = _fjsp_setzero_v2r8();
 474             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 475             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 476             ewtabFn          = _fjsp_setzero_v2r8();
 477             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 478             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 479             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 480             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 481             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 482
 483             /* CUBIC SPLINE TABLE DISPERSION */
 484             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 485             F                = _fjsp_setzero_v2r8();
 486             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 487             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 488             H                = _fjsp_setzero_v2r8();
 489             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 490             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 491             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 492             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 493             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 494             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 495
 496             /* CUBIC SPLINE TABLE REPULSION */
 497             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 498             F                = _fjsp_setzero_v2r8();
 499             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 500             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 501             H                = _fjsp_setzero_v2r8();
 502             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 503             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 504             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 505             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 506             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 507             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 508             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 509             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 510
 511             /* Update potential sum for this i atom from the interaction with this j atom. */
 512             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 513             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 514             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 515             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 516
 517             fscal            = _fjsp_add_v2r8(felec,fvdw);
 518
 519             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 520
 521             /* Update vectorial force */
 522             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 523             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 524             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 525
 526             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 527             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 528             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 529
 530             /**************************
 531              * CALCULATE INTERACTIONS *
 532              **************************/
 533
 534             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 535
 536             /* Compute parameters for interactions between i and j atoms */
 537             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 538
 539             /* EWALD ELECTROSTATICS */
 540
 541             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 542             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 543             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 544             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 545             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 546
 547             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 548             ewtabD           = _fjsp_setzero_v2r8();
 549             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 550             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 551             ewtabFn          = _fjsp_setzero_v2r8();
 552             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 553             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 554             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 555             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 556             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 557
 558             /* Update potential sum for this i atom from the interaction with this j atom. */
 559             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 560             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 561
 562             fscal            = felec;
 563
 564             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 565
 566             /* Update vectorial force */
 567             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 568             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 569             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 570
 571             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 572             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 573             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 574
 575             /**************************
 576              * CALCULATE INTERACTIONS *
 577              **************************/
 578
 579             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 580
 581             /* Compute parameters for interactions between i and j atoms */
 582             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 583
 584             /* EWALD ELECTROSTATICS */
 585
 586             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 587             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 588             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 589             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 590             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 591
 592             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 593             ewtabD           = _fjsp_setzero_v2r8();
 594             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 595             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 596             ewtabFn          = _fjsp_setzero_v2r8();
 597             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 598             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 599             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 600             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 601             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 602
 603             /* Update potential sum for this i atom from the interaction with this j atom. */
 604             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 605             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 606
 607             fscal            = felec;
 608
 609             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 610
 611             /* Update vectorial force */
 612             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 613             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 614             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 615
 616             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 617             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 618             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 619
 620             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 621
 622             /* Inner loop uses 169 flops */
 623         }
 624
 625         /* End of innermost loop */
 626
 627         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 628                                               f+i_coord_offset,fshift+i_shift_offset);
 629
 630         ggid                        = gid[iidx];
 631         /* Update potential energies */
 632         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 633         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 634
 635         /* Increment number of inner iterations */
 636         inneriter                  += j_index_end - j_index_start;
 637
 638         /* Outer loop uses 20 flops */
 639     }
 640
 641     /* Increment number of outer iterations */
 642     outeriter        += nri;
 643
 644     /* Update outer/inner flops */
 645
 646     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*169);
 647 }
 648 /*
 649  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 650  * Electrostatics interaction: Ewald
 651  * VdW interaction:            CubicSplineTable
 652  * Geometry:                   Water3-Particle
 653  * Calculate force/pot:        Force
 654  */
 655 void
 656 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 657                     (t_nblist                    * gmx_restrict       nlist,
 658                      rvec                        * gmx_restrict          xx,
 659                      rvec                        * gmx_restrict          ff,
 660                      t_forcerec                  * gmx_restrict          fr,
 661                      t_mdatoms                   * gmx_restrict     mdatoms,
 662                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 663                      t_nrnb                      * gmx_restrict        nrnb)
 664 {
 665     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 666      * just 0 for non-waters.
 667      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 668      * jnr indices corresponding to data put in the four positions in the SIMD register.
 669      */
 670     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 671     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 672     int              jnrA,jnrB;
 673     int              j_coord_offsetA,j_coord_offsetB;
 674     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 675     real             rcutoff_scalar;
 676     real             *shiftvec,*fshift,*x,*f;
 677     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 678     int              vdwioffset0;
 679     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 680     int              vdwioffset1;
 681     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 682     int              vdwioffset2;
 683     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 684     int              vdwjidx0A,vdwjidx0B;
 685     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 686     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 687     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 688     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 689     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 690     real             *charge;
 691     int              nvdwtype;
 692     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 693     int              *vdwtype;
 694     real             *vdwparam;
 695     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 696     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 697     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 698     real             *vftab;
 699     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 700     real             *ewtab;
 701     _fjsp_v2r8       itab_tmp;
 702     _fjsp_v2r8       dummy_mask,cutoff_mask;
 703     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 704     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 705     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 706
 707     x                = xx[0];
 708     f                = ff[0];
 709
 710     nri              = nlist->nri;
 711     iinr             = nlist->iinr;
 712     jindex           = nlist->jindex;
 713     jjnr             = nlist->jjnr;
 714     shiftidx         = nlist->shift;
 715     gid              = nlist->gid;
 716     shiftvec         = fr->shift_vec[0];
 717     fshift           = fr->fshift[0];
 718     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 719     charge           = mdatoms->chargeA;
 720     nvdwtype         = fr->ntype;
 721     vdwparam         = fr->nbfp;
 722     vdwtype          = mdatoms->typeA;
 723
 724     vftab            = kernel_data->table_vdw->data;
 725     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 726
 727     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 728     ewtab            = fr->ic->tabq_coul_F;
 729     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 730     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 731
 732     /* Setup water-specific parameters */
 733     inr              = nlist->iinr[0];
 734     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 735     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 736     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 737     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 738
 739     /* Avoid stupid compiler warnings */
 740     jnrA = jnrB = 0;
 741     j_coord_offsetA = 0;
 742     j_coord_offsetB = 0;
 743
 744     outeriter        = 0;
 745     inneriter        = 0;
 746
 747     /* Start outer loop over neighborlists */
 748     for(iidx=0; iidx<nri; iidx++)
 749     {
 750         /* Load shift vector for this list */
 751         i_shift_offset   = DIM*shiftidx[iidx];
 752
 753         /* Load limits for loop over neighbors */
 754         j_index_start    = jindex[iidx];
 755         j_index_end      = jindex[iidx+1];
 756
 757         /* Get outer coordinate index */
 758         inr              = iinr[iidx];
 759         i_coord_offset   = DIM*inr;
 760
 761         /* Load i particle coords and add shift vector */
 762         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 763                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 764
 765         fix0             = _fjsp_setzero_v2r8();
 766         fiy0             = _fjsp_setzero_v2r8();
 767         fiz0             = _fjsp_setzero_v2r8();
 768         fix1             = _fjsp_setzero_v2r8();
 769         fiy1             = _fjsp_setzero_v2r8();
 770         fiz1             = _fjsp_setzero_v2r8();
 771         fix2             = _fjsp_setzero_v2r8();
 772         fiy2             = _fjsp_setzero_v2r8();
 773         fiz2             = _fjsp_setzero_v2r8();
 774
 775         /* Start inner kernel loop */
 776         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 777         {
 778
 779             /* Get j neighbor index, and coordinate index */
 780             jnrA             = jjnr[jidx];
 781             jnrB             = jjnr[jidx+1];
 782             j_coord_offsetA  = DIM*jnrA;
 783             j_coord_offsetB  = DIM*jnrB;
 784
 785             /* load j atom coordinates */
 786             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 787                                               &jx0,&jy0,&jz0);
 788
 789             /* Calculate displacement vector */
 790             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 791             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 792             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 793             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 794             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 795             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 796             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 797             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 798             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 799
 800             /* Calculate squared distance and things based on it */
 801             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 802             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 803             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 804
 805             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 806             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 807             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 808
 809             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 810             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 811             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 812
 813             /* Load parameters for j particles */
 814             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 815             vdwjidx0A        = 2*vdwtype[jnrA+0];
 816             vdwjidx0B        = 2*vdwtype[jnrB+0];
 817
 818             fjx0             = _fjsp_setzero_v2r8();
 819             fjy0             = _fjsp_setzero_v2r8();
 820             fjz0             = _fjsp_setzero_v2r8();
 821
 822             /**************************
 823              * CALCULATE INTERACTIONS *
 824              **************************/
 825
 826             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 827
 828             /* Compute parameters for interactions between i and j atoms */
 829             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 830             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 831                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 832
 833             /* Calculate table index by multiplying r with table scale and truncate to integer */
 834             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 835             itab_tmp         = _fjsp_dtox_v2r8(rt);
 836             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 837             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 838             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 839
 840             vfconv.i[0]     *= 8;
 841             vfconv.i[1]     *= 8;
 842
 843             /* EWALD ELECTROSTATICS */
 844
 845             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 846             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 847             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 848             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 849             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 850
 851             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 852                                          &ewtabF,&ewtabFn);
 853             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 854             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 855
 856             /* CUBIC SPLINE TABLE DISPERSION */
 857             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 858             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 859             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 860             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 861             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 862             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 863             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 864             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 865             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 866
 867             /* CUBIC SPLINE TABLE REPULSION */
 868             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 869             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 870             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 871             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 872             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 873             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 874             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 875             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 876             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 877             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 878
 879             fscal            = _fjsp_add_v2r8(felec,fvdw);
 880
 881             /* Update vectorial force */
 882             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 883             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 884             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 885
 886             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 887             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 888             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 889
 890             /**************************
 891              * CALCULATE INTERACTIONS *
 892              **************************/
 893
 894             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 895
 896             /* Compute parameters for interactions between i and j atoms */
 897             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 898
 899             /* EWALD ELECTROSTATICS */
 900
 901             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 902             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 903             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 904             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 905             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 906
 907             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 908                                          &ewtabF,&ewtabFn);
 909             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 910             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 911
 912             fscal            = felec;
 913
 914             /* Update vectorial force */
 915             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 916             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 917             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 918
 919             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 920             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 921             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 922
 923             /**************************
 924              * CALCULATE INTERACTIONS *
 925              **************************/
 926
 927             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 928
 929             /* Compute parameters for interactions between i and j atoms */
 930             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 931
 932             /* EWALD ELECTROSTATICS */
 933
 934             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 935             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 936             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 937             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 938             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 939
 940             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 941                                          &ewtabF,&ewtabFn);
 942             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 943             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 944
 945             fscal            = felec;
 946
 947             /* Update vectorial force */
 948             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 949             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 950             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 951
 952             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 953             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 954             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 955
 956             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 957
 958             /* Inner loop uses 146 flops */
 959         }
 960
 961         if(jidx<j_index_end)
 962         {
 963
 964             jnrA             = jjnr[jidx];
 965             j_coord_offsetA  = DIM*jnrA;
 966
 967             /* load j atom coordinates */
 968             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 969                                               &jx0,&jy0,&jz0);
 970
 971             /* Calculate displacement vector */
 972             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 973             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 974             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 975             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 976             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 977             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 978             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 979             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 980             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 981
 982             /* Calculate squared distance and things based on it */
 983             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 984             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 985             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 986
 987             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 988             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 989             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 990
 991             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 992             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 993             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 994
 995             /* Load parameters for j particles */
 996             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 997             vdwjidx0A        = 2*vdwtype[jnrA+0];
 998
 999             fjx0             = _fjsp_setzero_v2r8();
1000             fjy0             = _fjsp_setzero_v2r8();
1001             fjz0             = _fjsp_setzero_v2r8();
1002
1003             /**************************
1004              * CALCULATE INTERACTIONS *
1005              **************************/
1006
1007             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1008
1009             /* Compute parameters for interactions between i and j atoms */
1010             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1011             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
1012                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
1013
1014             /* Calculate table index by multiplying r with table scale and truncate to integer */
1015             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1016             itab_tmp         = _fjsp_dtox_v2r8(rt);
1017             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1018             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1019             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1020
1021             vfconv.i[0]     *= 8;
1022             vfconv.i[1]     *= 8;
1023
1024             /* EWALD ELECTROSTATICS */
1025
1026             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1027             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1028             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1029             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1030             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1031
1032             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1033             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1034             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1035
1036             /* CUBIC SPLINE TABLE DISPERSION */
1037             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1038             F                = _fjsp_setzero_v2r8();
1039             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1040             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1041             H                = _fjsp_setzero_v2r8();
1042             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1043             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1044             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1045             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1046
1047             /* CUBIC SPLINE TABLE REPULSION */
1048             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1049             F                = _fjsp_setzero_v2r8();
1050             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1051             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1052             H                = _fjsp_setzero_v2r8();
1053             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1054             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1055             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1056             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1057             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1058
1059             fscal            = _fjsp_add_v2r8(felec,fvdw);
1060
1061             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1062
1063             /* Update vectorial force */
1064             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1065             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1066             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1067
1068             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1069             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1070             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1071
1072             /**************************
1073              * CALCULATE INTERACTIONS *
1074              **************************/
1075
1076             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1077
1078             /* Compute parameters for interactions between i and j atoms */
1079             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1080
1081             /* EWALD ELECTROSTATICS */
1082
1083             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1084             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1085             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1086             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1087             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1088
1089             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1090             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1091             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1092
1093             fscal            = felec;
1094
1095             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1096
1097             /* Update vectorial force */
1098             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1099             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1100             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1101
1102             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1103             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1104             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1105
1106             /**************************
1107              * CALCULATE INTERACTIONS *
1108              **************************/
1109
1110             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1111
1112             /* Compute parameters for interactions between i and j atoms */
1113             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1114
1115             /* EWALD ELECTROSTATICS */
1116
1117             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1118             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1119             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1120             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1121             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1122
1123             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1124             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1125             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1126
1127             fscal            = felec;
1128
1129             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1130
1131             /* Update vectorial force */
1132             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1133             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1134             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1135
1136             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1137             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1138             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1139
1140             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1141
1142             /* Inner loop uses 146 flops */
1143         }
1144
1145         /* End of innermost loop */
1146
1147         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1148                                               f+i_coord_offset,fshift+i_shift_offset);
1149
1150         /* Increment number of inner iterations */
1151         inneriter                  += j_index_end - j_index_start;
1152
1153         /* Outer loop uses 18 flops */
1154     }
1155
1156     /* Increment number of outer iterations */
1157     outeriter        += nri;
1158
1159     /* Update outer/inner flops */
1160
1161     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*146);
1162 }