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 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "gromacs/legacyheaders/types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "gromacs/legacyheaders/nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water3-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_VF_sparc64_hpc_ace_double
  58                     (t_nblist                    * gmx_restrict       nlist,
  59                      rvec                        * gmx_restrict          xx,
  60                      rvec                        * gmx_restrict          ff,
  61                      t_forcerec                  * gmx_restrict          fr,
  62                      t_mdatoms                   * gmx_restrict     mdatoms,
  63                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  64                      t_nrnb                      * gmx_restrict        nrnb)
  65 {
  66     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  67      * just 0 for non-waters.
  68      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  69      * jnr indices corresponding to data put in the four positions in the SIMD register.
  70      */
  71     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  72     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  73     int              jnrA,jnrB;
  74     int              j_coord_offsetA,j_coord_offsetB;
  75     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  76     real             rcutoff_scalar;
  77     real             *shiftvec,*fshift,*x,*f;
  78     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  79     int              vdwioffset0;
  80     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  81     int              vdwioffset1;
  82     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
  83     int              vdwioffset2;
  84     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
  85     int              vdwjidx0A,vdwjidx0B;
  86     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  87     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  88     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
  89     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
  90     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  91     real             *charge;
  92     int              nvdwtype;
  93     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  94     int              *vdwtype;
  95     real             *vdwparam;
  96     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  97     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  98     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  99     real             *vftab;
 100     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 101     real             *ewtab;
 102     _fjsp_v2r8       itab_tmp;
 103     _fjsp_v2r8       dummy_mask,cutoff_mask;
 104     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 105     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 106     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 107
 108     x                = xx[0];
 109     f                = ff[0];
 110
 111     nri              = nlist->nri;
 112     iinr             = nlist->iinr;
 113     jindex           = nlist->jindex;
 114     jjnr             = nlist->jjnr;
 115     shiftidx         = nlist->shift;
 116     gid              = nlist->gid;
 117     shiftvec         = fr->shift_vec[0];
 118     fshift           = fr->fshift[0];
 119     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 120     charge           = mdatoms->chargeA;
 121     nvdwtype         = fr->ntype;
 122     vdwparam         = fr->nbfp;
 123     vdwtype          = mdatoms->typeA;
 124
 125     vftab            = kernel_data->table_vdw->data;
 126     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 127
 128     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 129     ewtab            = fr->ic->tabq_coul_FDV0;
 130     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 131     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 132
 133     /* Setup water-specific parameters */
 134     inr              = nlist->iinr[0];
 135     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 136     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 137     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 138     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 139
 140     /* Avoid stupid compiler warnings */
 141     jnrA = jnrB = 0;
 142     j_coord_offsetA = 0;
 143     j_coord_offsetB = 0;
 144
 145     outeriter        = 0;
 146     inneriter        = 0;
 147
 148     /* Start outer loop over neighborlists */
 149     for(iidx=0; iidx<nri; iidx++)
 150     {
 151         /* Load shift vector for this list */
 152         i_shift_offset   = DIM*shiftidx[iidx];
 153
 154         /* Load limits for loop over neighbors */
 155         j_index_start    = jindex[iidx];
 156         j_index_end      = jindex[iidx+1];
 157
 158         /* Get outer coordinate index */
 159         inr              = iinr[iidx];
 160         i_coord_offset   = DIM*inr;
 161
 162         /* Load i particle coords and add shift vector */
 163         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 164                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 165
 166         fix0             = _fjsp_setzero_v2r8();
 167         fiy0             = _fjsp_setzero_v2r8();
 168         fiz0             = _fjsp_setzero_v2r8();
 169         fix1             = _fjsp_setzero_v2r8();
 170         fiy1             = _fjsp_setzero_v2r8();
 171         fiz1             = _fjsp_setzero_v2r8();
 172         fix2             = _fjsp_setzero_v2r8();
 173         fiy2             = _fjsp_setzero_v2r8();
 174         fiz2             = _fjsp_setzero_v2r8();
 175
 176         /* Reset potential sums */
 177         velecsum         = _fjsp_setzero_v2r8();
 178         vvdwsum          = _fjsp_setzero_v2r8();
 179
 180         /* Start inner kernel loop */
 181         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 182         {
 183
 184             /* Get j neighbor index, and coordinate index */
 185             jnrA             = jjnr[jidx];
 186             jnrB             = jjnr[jidx+1];
 187             j_coord_offsetA  = DIM*jnrA;
 188             j_coord_offsetB  = DIM*jnrB;
 189
 190             /* load j atom coordinates */
 191             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 192                                               &jx0,&jy0,&jz0);
 193
 194             /* Calculate displacement vector */
 195             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 196             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 197             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 198             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 199             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 200             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 201             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 202             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 203             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 204
 205             /* Calculate squared distance and things based on it */
 206             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 207             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 208             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 209
 210             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 211             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 212             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 213
 214             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 215             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 216             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 217
 218             /* Load parameters for j particles */
 219             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 220             vdwjidx0A        = 2*vdwtype[jnrA+0];
 221             vdwjidx0B        = 2*vdwtype[jnrB+0];
 222
 223             fjx0             = _fjsp_setzero_v2r8();
 224             fjy0             = _fjsp_setzero_v2r8();
 225             fjz0             = _fjsp_setzero_v2r8();
 226
 227             /**************************
 228              * CALCULATE INTERACTIONS *
 229              **************************/
 230
 231             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 232
 233             /* Compute parameters for interactions between i and j atoms */
 234             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 235             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 236                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 237
 238             /* Calculate table index by multiplying r with table scale and truncate to integer */
 239             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 240             itab_tmp         = _fjsp_dtox_v2r8(rt);
 241             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 242             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 243             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 244
 245             vfconv.i[0]     *= 8;
 246             vfconv.i[1]     *= 8;
 247
 248             /* EWALD ELECTROSTATICS */
 249
 250             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 251             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 252             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 253             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 254             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 255
 256             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 257             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 258             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 259             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 260             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 261             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 262             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 263             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 264             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 265             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 266
 267             /* CUBIC SPLINE TABLE DISPERSION */
 268             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 269             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 270             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 271             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 272             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 273             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 274             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 275             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 276             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 277             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 278             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 279
 280             /* CUBIC SPLINE TABLE REPULSION */
 281             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 282             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 283             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 284             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 285             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 286             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 287             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 288             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 289             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 290             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 291             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 292             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 293             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 294
 295             /* Update potential sum for this i atom from the interaction with this j atom. */
 296             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 297             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 298
 299             fscal            = _fjsp_add_v2r8(felec,fvdw);
 300
 301             /* Update vectorial force */
 302             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 303             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 304             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 305
 306             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 307             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 308             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 309
 310             /**************************
 311              * CALCULATE INTERACTIONS *
 312              **************************/
 313
 314             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 315
 316             /* Compute parameters for interactions between i and j atoms */
 317             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 318
 319             /* EWALD ELECTROSTATICS */
 320
 321             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 322             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 323             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 324             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 325             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 326
 327             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 328             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 329             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 330             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 331             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 332             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 333             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 334             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 335             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 336             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 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             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 345             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 346             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 347
 348             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 349             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 350             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 351
 352             /**************************
 353              * CALCULATE INTERACTIONS *
 354              **************************/
 355
 356             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 357
 358             /* Compute parameters for interactions between i and j atoms */
 359             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 360
 361             /* EWALD ELECTROSTATICS */
 362
 363             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 364             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 365             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 366             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 367             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 368
 369             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 370             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 371             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 372             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 373             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 374             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 375             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 376             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 377             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 378             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 379
 380             /* Update potential sum for this i atom from the interaction with this j atom. */
 381             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 382
 383             fscal            = felec;
 384
 385             /* Update vectorial force */
 386             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 387             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 388             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 389
 390             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 391             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 392             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 393
 394             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 395
 396             /* Inner loop uses 169 flops */
 397         }
 398
 399         if(jidx<j_index_end)
 400         {
 401
 402             jnrA             = jjnr[jidx];
 403             j_coord_offsetA  = DIM*jnrA;
 404
 405             /* load j atom coordinates */
 406             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 407                                               &jx0,&jy0,&jz0);
 408
 409             /* Calculate displacement vector */
 410             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 411             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 412             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 413             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 414             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 415             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 416             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 417             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 418             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 419
 420             /* Calculate squared distance and things based on it */
 421             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 422             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 423             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 424
 425             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 426             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 427             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 428
 429             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 430             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 431             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 432
 433             /* Load parameters for j particles */
 434             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 435             vdwjidx0A        = 2*vdwtype[jnrA+0];
 436
 437             fjx0             = _fjsp_setzero_v2r8();
 438             fjy0             = _fjsp_setzero_v2r8();
 439             fjz0             = _fjsp_setzero_v2r8();
 440
 441             /**************************
 442              * CALCULATE INTERACTIONS *
 443              **************************/
 444
 445             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 446
 447             /* Compute parameters for interactions between i and j atoms */
 448             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 449             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 450
 451             /* Calculate table index by multiplying r with table scale and truncate to integer */
 452             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 453             itab_tmp         = _fjsp_dtox_v2r8(rt);
 454             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 455             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 456             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 457
 458             vfconv.i[0]     *= 8;
 459             vfconv.i[1]     *= 8;
 460
 461             /* EWALD ELECTROSTATICS */
 462
 463             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 464             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 465             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 466             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 467             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 468
 469             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 470             ewtabD           = _fjsp_setzero_v2r8();
 471             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 472             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 473             ewtabFn          = _fjsp_setzero_v2r8();
 474             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 475             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 476             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 477             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 478             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 479
 480             /* CUBIC SPLINE TABLE DISPERSION */
 481             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 482             F                = _fjsp_setzero_v2r8();
 483             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 484             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 485             H                = _fjsp_setzero_v2r8();
 486             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 487             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 488             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 489             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 490             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 491             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 492
 493             /* CUBIC SPLINE TABLE REPULSION */
 494             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 495             F                = _fjsp_setzero_v2r8();
 496             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 497             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 498             H                = _fjsp_setzero_v2r8();
 499             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 500             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 501             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 502             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 503             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 504             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 505             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 506             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 507
 508             /* Update potential sum for this i atom from the interaction with this j atom. */
 509             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 510             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 511             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 512             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 513
 514             fscal            = _fjsp_add_v2r8(felec,fvdw);
 515
 516             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 517
 518             /* Update vectorial force */
 519             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 520             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 521             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 522
 523             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 524             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 525             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 526
 527             /**************************
 528              * CALCULATE INTERACTIONS *
 529              **************************/
 530
 531             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 532
 533             /* Compute parameters for interactions between i and j atoms */
 534             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 535
 536             /* EWALD ELECTROSTATICS */
 537
 538             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 539             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 540             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 541             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 542             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 543
 544             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 545             ewtabD           = _fjsp_setzero_v2r8();
 546             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 547             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 548             ewtabFn          = _fjsp_setzero_v2r8();
 549             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 550             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 551             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 552             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 553             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 554
 555             /* Update potential sum for this i atom from the interaction with this j atom. */
 556             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 557             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 558
 559             fscal            = felec;
 560
 561             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 562
 563             /* Update vectorial force */
 564             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 565             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 566             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 567
 568             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 569             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 570             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 571
 572             /**************************
 573              * CALCULATE INTERACTIONS *
 574              **************************/
 575
 576             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 577
 578             /* Compute parameters for interactions between i and j atoms */
 579             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 580
 581             /* EWALD ELECTROSTATICS */
 582
 583             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 584             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 585             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 586             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 587             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 588
 589             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 590             ewtabD           = _fjsp_setzero_v2r8();
 591             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 592             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 593             ewtabFn          = _fjsp_setzero_v2r8();
 594             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 595             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 596             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 597             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 598             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 599
 600             /* Update potential sum for this i atom from the interaction with this j atom. */
 601             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 602             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 603
 604             fscal            = felec;
 605
 606             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 607
 608             /* Update vectorial force */
 609             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 610             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 611             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 612
 613             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 614             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 615             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 616
 617             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 618
 619             /* Inner loop uses 169 flops */
 620         }
 621
 622         /* End of innermost loop */
 623
 624         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
 625                                               f+i_coord_offset,fshift+i_shift_offset);
 626
 627         ggid                        = gid[iidx];
 628         /* Update potential energies */
 629         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 630         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 631
 632         /* Increment number of inner iterations */
 633         inneriter                  += j_index_end - j_index_start;
 634
 635         /* Outer loop uses 20 flops */
 636     }
 637
 638     /* Increment number of outer iterations */
 639     outeriter        += nri;
 640
 641     /* Update outer/inner flops */
 642
 643     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*169);
 644 }
 645 /*
 646  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 647  * Electrostatics interaction: Ewald
 648  * VdW interaction:            CubicSplineTable
 649  * Geometry:                   Water3-Particle
 650  * Calculate force/pot:        Force
 651  */
 652 void
 653 nb_kernel_ElecEw_VdwCSTab_GeomW3P1_F_sparc64_hpc_ace_double
 654                     (t_nblist                    * gmx_restrict       nlist,
 655                      rvec                        * gmx_restrict          xx,
 656                      rvec                        * gmx_restrict          ff,
 657                      t_forcerec                  * gmx_restrict          fr,
 658                      t_mdatoms                   * gmx_restrict     mdatoms,
 659                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 660                      t_nrnb                      * gmx_restrict        nrnb)
 661 {
 662     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 663      * just 0 for non-waters.
 664      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 665      * jnr indices corresponding to data put in the four positions in the SIMD register.
 666      */
 667     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 668     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 669     int              jnrA,jnrB;
 670     int              j_coord_offsetA,j_coord_offsetB;
 671     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 672     real             rcutoff_scalar;
 673     real             *shiftvec,*fshift,*x,*f;
 674     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 675     int              vdwioffset0;
 676     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 677     int              vdwioffset1;
 678     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 679     int              vdwioffset2;
 680     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 681     int              vdwjidx0A,vdwjidx0B;
 682     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 683     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 684     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 685     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 686     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 687     real             *charge;
 688     int              nvdwtype;
 689     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 690     int              *vdwtype;
 691     real             *vdwparam;
 692     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 693     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 694     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 695     real             *vftab;
 696     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 697     real             *ewtab;
 698     _fjsp_v2r8       itab_tmp;
 699     _fjsp_v2r8       dummy_mask,cutoff_mask;
 700     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 701     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 702     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 703
 704     x                = xx[0];
 705     f                = ff[0];
 706
 707     nri              = nlist->nri;
 708     iinr             = nlist->iinr;
 709     jindex           = nlist->jindex;
 710     jjnr             = nlist->jjnr;
 711     shiftidx         = nlist->shift;
 712     gid              = nlist->gid;
 713     shiftvec         = fr->shift_vec[0];
 714     fshift           = fr->fshift[0];
 715     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 716     charge           = mdatoms->chargeA;
 717     nvdwtype         = fr->ntype;
 718     vdwparam         = fr->nbfp;
 719     vdwtype          = mdatoms->typeA;
 720
 721     vftab            = kernel_data->table_vdw->data;
 722     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 723
 724     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 725     ewtab            = fr->ic->tabq_coul_F;
 726     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 727     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 728
 729     /* Setup water-specific parameters */
 730     inr              = nlist->iinr[0];
 731     iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+0]));
 732     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 733     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 734     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 735
 736     /* Avoid stupid compiler warnings */
 737     jnrA = jnrB = 0;
 738     j_coord_offsetA = 0;
 739     j_coord_offsetB = 0;
 740
 741     outeriter        = 0;
 742     inneriter        = 0;
 743
 744     /* Start outer loop over neighborlists */
 745     for(iidx=0; iidx<nri; iidx++)
 746     {
 747         /* Load shift vector for this list */
 748         i_shift_offset   = DIM*shiftidx[iidx];
 749
 750         /* Load limits for loop over neighbors */
 751         j_index_start    = jindex[iidx];
 752         j_index_end      = jindex[iidx+1];
 753
 754         /* Get outer coordinate index */
 755         inr              = iinr[iidx];
 756         i_coord_offset   = DIM*inr;
 757
 758         /* Load i particle coords and add shift vector */
 759         gmx_fjsp_load_shift_and_3rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 760                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
 761
 762         fix0             = _fjsp_setzero_v2r8();
 763         fiy0             = _fjsp_setzero_v2r8();
 764         fiz0             = _fjsp_setzero_v2r8();
 765         fix1             = _fjsp_setzero_v2r8();
 766         fiy1             = _fjsp_setzero_v2r8();
 767         fiz1             = _fjsp_setzero_v2r8();
 768         fix2             = _fjsp_setzero_v2r8();
 769         fiy2             = _fjsp_setzero_v2r8();
 770         fiz2             = _fjsp_setzero_v2r8();
 771
 772         /* Start inner kernel loop */
 773         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 774         {
 775
 776             /* Get j neighbor index, and coordinate index */
 777             jnrA             = jjnr[jidx];
 778             jnrB             = jjnr[jidx+1];
 779             j_coord_offsetA  = DIM*jnrA;
 780             j_coord_offsetB  = DIM*jnrB;
 781
 782             /* load j atom coordinates */
 783             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 784                                               &jx0,&jy0,&jz0);
 785
 786             /* Calculate displacement vector */
 787             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 788             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 789             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 790             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 791             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 792             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 793             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 794             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 795             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 796
 797             /* Calculate squared distance and things based on it */
 798             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 799             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 800             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 801
 802             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 803             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 804             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 805
 806             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 807             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 808             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 809
 810             /* Load parameters for j particles */
 811             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 812             vdwjidx0A        = 2*vdwtype[jnrA+0];
 813             vdwjidx0B        = 2*vdwtype[jnrB+0];
 814
 815             fjx0             = _fjsp_setzero_v2r8();
 816             fjy0             = _fjsp_setzero_v2r8();
 817             fjz0             = _fjsp_setzero_v2r8();
 818
 819             /**************************
 820              * CALCULATE INTERACTIONS *
 821              **************************/
 822
 823             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 824
 825             /* Compute parameters for interactions between i and j atoms */
 826             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 827             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 828                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 829
 830             /* Calculate table index by multiplying r with table scale and truncate to integer */
 831             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 832             itab_tmp         = _fjsp_dtox_v2r8(rt);
 833             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 834             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 835             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 836
 837             vfconv.i[0]     *= 8;
 838             vfconv.i[1]     *= 8;
 839
 840             /* EWALD ELECTROSTATICS */
 841
 842             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 843             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 844             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 845             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 846             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 847
 848             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 849                                          &ewtabF,&ewtabFn);
 850             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 851             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 852
 853             /* CUBIC SPLINE TABLE DISPERSION */
 854             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 855             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 856             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 857             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 858             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 859             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 860             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 861             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 862             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 863
 864             /* CUBIC SPLINE TABLE REPULSION */
 865             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 866             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 867             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 868             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 869             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 870             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 871             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 872             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 873             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 874             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 875
 876             fscal            = _fjsp_add_v2r8(felec,fvdw);
 877
 878             /* Update vectorial force */
 879             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 880             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 881             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 882
 883             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 884             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 885             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 886
 887             /**************************
 888              * CALCULATE INTERACTIONS *
 889              **************************/
 890
 891             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 892
 893             /* Compute parameters for interactions between i and j atoms */
 894             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 895
 896             /* EWALD ELECTROSTATICS */
 897
 898             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 899             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 900             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 901             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 902             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 903
 904             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 905                                          &ewtabF,&ewtabFn);
 906             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 907             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 908
 909             fscal            = felec;
 910
 911             /* Update vectorial force */
 912             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 913             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 914             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 915
 916             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 917             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 918             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 919
 920             /**************************
 921              * CALCULATE INTERACTIONS *
 922              **************************/
 923
 924             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 925
 926             /* Compute parameters for interactions between i and j atoms */
 927             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 928
 929             /* EWALD ELECTROSTATICS */
 930
 931             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 932             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 933             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 934             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 935             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 936
 937             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 938                                          &ewtabF,&ewtabFn);
 939             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 940             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 941
 942             fscal            = felec;
 943
 944             /* Update vectorial force */
 945             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 946             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 947             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 948
 949             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 950             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 951             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 952
 953             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 954
 955             /* Inner loop uses 146 flops */
 956         }
 957
 958         if(jidx<j_index_end)
 959         {
 960
 961             jnrA             = jjnr[jidx];
 962             j_coord_offsetA  = DIM*jnrA;
 963
 964             /* load j atom coordinates */
 965             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 966                                               &jx0,&jy0,&jz0);
 967
 968             /* Calculate displacement vector */
 969             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 970             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 971             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 972             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 973             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 974             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 975             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 976             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 977             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 978
 979             /* Calculate squared distance and things based on it */
 980             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 981             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 982             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 983
 984             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 985             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 986             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 987
 988             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 989             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 990             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 991
 992             /* Load parameters for j particles */
 993             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 994             vdwjidx0A        = 2*vdwtype[jnrA+0];
 995
 996             fjx0             = _fjsp_setzero_v2r8();
 997             fjy0             = _fjsp_setzero_v2r8();
 998             fjz0             = _fjsp_setzero_v2r8();
 999
1000             /**************************
1001              * CALCULATE INTERACTIONS *
1002              **************************/
1003
1004             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1005
1006             /* Compute parameters for interactions between i and j atoms */
1007             qq00             = _fjsp_mul_v2r8(iq0,jq0);
1008             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1009
1010             /* Calculate table index by multiplying r with table scale and truncate to integer */
1011             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1012             itab_tmp         = _fjsp_dtox_v2r8(rt);
1013             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1014             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1015             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1016
1017             vfconv.i[0]     *= 8;
1018             vfconv.i[1]     *= 8;
1019
1020             /* EWALD ELECTROSTATICS */
1021
1022             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1023             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
1024             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1025             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1026             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1027
1028             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1029             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1030             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
1031
1032             /* CUBIC SPLINE TABLE DISPERSION */
1033             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1034             F                = _fjsp_setzero_v2r8();
1035             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1036             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1037             H                = _fjsp_setzero_v2r8();
1038             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1039             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1040             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1041             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1042
1043             /* CUBIC SPLINE TABLE REPULSION */
1044             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1045             F                = _fjsp_setzero_v2r8();
1046             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1047             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1048             H                = _fjsp_setzero_v2r8();
1049             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1050             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1051             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1052             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1053             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1054
1055             fscal            = _fjsp_add_v2r8(felec,fvdw);
1056
1057             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1058
1059             /* Update vectorial force */
1060             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1061             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1062             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1063
1064             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1065             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1066             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1067
1068             /**************************
1069              * CALCULATE INTERACTIONS *
1070              **************************/
1071
1072             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1073
1074             /* Compute parameters for interactions between i and j atoms */
1075             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1076
1077             /* EWALD ELECTROSTATICS */
1078
1079             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1080             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1081             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1082             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1083             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1084
1085             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1086             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1087             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1088
1089             fscal            = felec;
1090
1091             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1092
1093             /* Update vectorial force */
1094             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1095             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1096             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1097
1098             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1099             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1100             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1101
1102             /**************************
1103              * CALCULATE INTERACTIONS *
1104              **************************/
1105
1106             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1107
1108             /* Compute parameters for interactions between i and j atoms */
1109             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1110
1111             /* EWALD ELECTROSTATICS */
1112
1113             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1114             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1115             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1116             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1117             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1118
1119             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1120             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1121             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1122
1123             fscal            = felec;
1124
1125             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1126
1127             /* Update vectorial force */
1128             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1129             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1130             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1131
1132             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1133             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1134             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1135
1136             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1137
1138             /* Inner loop uses 146 flops */
1139         }
1140
1141         /* End of innermost loop */
1142
1143         gmx_fjsp_update_iforce_3atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
1144                                               f+i_coord_offset,fshift+i_shift_offset);
1145
1146         /* Increment number of inner iterations */
1147         inneriter                  += j_index_end - j_index_start;
1148
1149         /* Outer loop uses 18 flops */
1150     }
1151
1152     /* Increment number of outer iterations */
1153     outeriter        += nri;
1154
1155     /* Update outer/inner flops */
1156
1157     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*146);
1158 }