src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_VdwLJ_GeomW4P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
   5  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   6  * and including many others, as listed in the AUTHORS file in the
   7  * top-level source directory and at http://www.gromacs.org.
   8  *
   9  * GROMACS is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public License
  11  * as published by the Free Software Foundation; either version 2.1
  12  * of the License, or (at your option) any later version.
  13  *
  14  * GROMACS is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with GROMACS; if not, see
  21  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  22  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  23  *
  24  * If you want to redistribute modifications to GROMACS, please
  25  * consider that scientific software is very special. Version
  26  * control is crucial - bugs must be traceable. We will be happy to
  27  * consider code for inclusion in the official distribution, but
  28  * derived work must not be called official GROMACS. Details are found
  29  * in the README & COPYING files - if they are missing, get the
  30  * official version at http://www.gromacs.org.
  31  *
  32  * To help us fund GROMACS development, we humbly ask that you cite
  33  * the research papers on the package. Check out http://www.gromacs.org.
  34  */
  35 /*
  36  * Note: this file was generated by the GROMACS sparc64_hpc_ace_double kernel generator.
  37  */
  38 #include "config.h"
  39
  40 #include <math.h>
  41
  42 #include "../nb_kernel.h"
  43 #include "types/simple.h"
  44 #include "gromacs/math/vec.h"
  45 #include "nrnb.h"
  46
  47 #include "kernelutil_sparc64_hpc_ace_double.h"
  48
  49 /*
  50  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: Ewald
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecEw_VdwLJ_GeomW4P1_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              vdwioffset3;
  86     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
  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       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
  93     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  94     real             *charge;
  95     int              nvdwtype;
  96     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  97     int              *vdwtype;
  98     real             *vdwparam;
  99     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 100     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 101     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 102     real             *ewtab;
 103     _fjsp_v2r8       itab_tmp;
 104     _fjsp_v2r8       dummy_mask,cutoff_mask;
 105     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 106     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 107     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 108
 109     x                = xx[0];
 110     f                = ff[0];
 111
 112     nri              = nlist->nri;
 113     iinr             = nlist->iinr;
 114     jindex           = nlist->jindex;
 115     jjnr             = nlist->jjnr;
 116     shiftidx         = nlist->shift;
 117     gid              = nlist->gid;
 118     shiftvec         = fr->shift_vec[0];
 119     fshift           = fr->fshift[0];
 120     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 121     charge           = mdatoms->chargeA;
 122     nvdwtype         = fr->ntype;
 123     vdwparam         = fr->nbfp;
 124     vdwtype          = mdatoms->typeA;
 125
 126     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 127     ewtab            = fr->ic->tabq_coul_FDV0;
 128     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 129     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 130
 131     /* Setup water-specific parameters */
 132     inr              = nlist->iinr[0];
 133     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 134     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 135     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 136     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 137
 138     /* Avoid stupid compiler warnings */
 139     jnrA = jnrB = 0;
 140     j_coord_offsetA = 0;
 141     j_coord_offsetB = 0;
 142
 143     outeriter        = 0;
 144     inneriter        = 0;
 145
 146     /* Start outer loop over neighborlists */
 147     for(iidx=0; iidx<nri; iidx++)
 148     {
 149         /* Load shift vector for this list */
 150         i_shift_offset   = DIM*shiftidx[iidx];
 151
 152         /* Load limits for loop over neighbors */
 153         j_index_start    = jindex[iidx];
 154         j_index_end      = jindex[iidx+1];
 155
 156         /* Get outer coordinate index */
 157         inr              = iinr[iidx];
 158         i_coord_offset   = DIM*inr;
 159
 160         /* Load i particle coords and add shift vector */
 161         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 162                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 163
 164         fix0             = _fjsp_setzero_v2r8();
 165         fiy0             = _fjsp_setzero_v2r8();
 166         fiz0             = _fjsp_setzero_v2r8();
 167         fix1             = _fjsp_setzero_v2r8();
 168         fiy1             = _fjsp_setzero_v2r8();
 169         fiz1             = _fjsp_setzero_v2r8();
 170         fix2             = _fjsp_setzero_v2r8();
 171         fiy2             = _fjsp_setzero_v2r8();
 172         fiz2             = _fjsp_setzero_v2r8();
 173         fix3             = _fjsp_setzero_v2r8();
 174         fiy3             = _fjsp_setzero_v2r8();
 175         fiz3             = _fjsp_setzero_v2r8();
 176
 177         /* Reset potential sums */
 178         velecsum         = _fjsp_setzero_v2r8();
 179         vvdwsum          = _fjsp_setzero_v2r8();
 180
 181         /* Start inner kernel loop */
 182         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 183         {
 184
 185             /* Get j neighbor index, and coordinate index */
 186             jnrA             = jjnr[jidx];
 187             jnrB             = jjnr[jidx+1];
 188             j_coord_offsetA  = DIM*jnrA;
 189             j_coord_offsetB  = DIM*jnrB;
 190
 191             /* load j atom coordinates */
 192             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 193                                               &jx0,&jy0,&jz0);
 194
 195             /* Calculate displacement vector */
 196             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 197             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 198             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 199             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 200             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 201             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 202             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 203             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 204             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 205             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 206             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 207             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 208
 209             /* Calculate squared distance and things based on it */
 210             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 211             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 212             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 213             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 214
 215             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 216             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 217             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 218
 219             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 220             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 221             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 222             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 223
 224             /* Load parameters for j particles */
 225             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 226             vdwjidx0A        = 2*vdwtype[jnrA+0];
 227             vdwjidx0B        = 2*vdwtype[jnrB+0];
 228
 229             fjx0             = _fjsp_setzero_v2r8();
 230             fjy0             = _fjsp_setzero_v2r8();
 231             fjz0             = _fjsp_setzero_v2r8();
 232
 233             /**************************
 234              * CALCULATE INTERACTIONS *
 235              **************************/
 236
 237             /* Compute parameters for interactions between i and j atoms */
 238             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 239                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 240
 241             /* LENNARD-JONES DISPERSION/REPULSION */
 242
 243             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 244             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 245             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 246             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 247             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 248
 249             /* Update potential sum for this i atom from the interaction with this j atom. */
 250             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 251
 252             fscal            = fvdw;
 253
 254             /* Update vectorial force */
 255             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 256             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 257             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 258
 259             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 260             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 261             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 262
 263             /**************************
 264              * CALCULATE INTERACTIONS *
 265              **************************/
 266
 267             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 268
 269             /* Compute parameters for interactions between i and j atoms */
 270             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 271
 272             /* EWALD ELECTROSTATICS */
 273
 274             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 275             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 276             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 277             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 278             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 279
 280             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 281             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 282             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 283             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 284             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 285             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 286             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 287             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 288             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 289             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 290
 291             /* Update potential sum for this i atom from the interaction with this j atom. */
 292             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 293
 294             fscal            = felec;
 295
 296             /* Update vectorial force */
 297             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 298             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 299             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 300
 301             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 302             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 303             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 304
 305             /**************************
 306              * CALCULATE INTERACTIONS *
 307              **************************/
 308
 309             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 310
 311             /* Compute parameters for interactions between i and j atoms */
 312             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 313
 314             /* EWALD ELECTROSTATICS */
 315
 316             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 317             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 318             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 319             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 320             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 321
 322             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 323             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 324             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 325             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 326             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 327             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 328             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 329             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 330             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 331             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 332
 333             /* Update potential sum for this i atom from the interaction with this j atom. */
 334             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 335
 336             fscal            = felec;
 337
 338             /* Update vectorial force */
 339             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 340             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 341             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 342
 343             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 344             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 345             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 346
 347             /**************************
 348              * CALCULATE INTERACTIONS *
 349              **************************/
 350
 351             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 352
 353             /* Compute parameters for interactions between i and j atoms */
 354             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 355
 356             /* EWALD ELECTROSTATICS */
 357
 358             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 359             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 360             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 361             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 362             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 363
 364             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 365             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 366             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 367             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 368             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 369             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 370             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 371             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 372             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 373             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 374
 375             /* Update potential sum for this i atom from the interaction with this j atom. */
 376             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 377
 378             fscal            = felec;
 379
 380             /* Update vectorial force */
 381             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 382             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 383             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 384
 385             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 386             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 387             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 388
 389             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 390
 391             /* Inner loop uses 170 flops */
 392         }
 393
 394         if(jidx<j_index_end)
 395         {
 396
 397             jnrA             = jjnr[jidx];
 398             j_coord_offsetA  = DIM*jnrA;
 399
 400             /* load j atom coordinates */
 401             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 402                                               &jx0,&jy0,&jz0);
 403
 404             /* Calculate displacement vector */
 405             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 406             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 407             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 408             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 409             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 410             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 411             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 412             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 413             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 414             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 415             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 416             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 417
 418             /* Calculate squared distance and things based on it */
 419             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 420             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 421             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 422             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 423
 424             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 425             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 426             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 427
 428             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 429             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 430             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 431             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 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             /* Compute parameters for interactions between i and j atoms */
 446             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 447
 448             /* LENNARD-JONES DISPERSION/REPULSION */
 449
 450             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 451             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 452             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 453             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 454             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 455
 456             /* Update potential sum for this i atom from the interaction with this j atom. */
 457             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 458             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 459
 460             fscal            = fvdw;
 461
 462             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 463
 464             /* Update vectorial force */
 465             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 466             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 467             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 468
 469             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 470             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 471             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 472
 473             /**************************
 474              * CALCULATE INTERACTIONS *
 475              **************************/
 476
 477             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 478
 479             /* Compute parameters for interactions between i and j atoms */
 480             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 481
 482             /* EWALD ELECTROSTATICS */
 483
 484             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 485             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 486             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 487             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 488             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 489
 490             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 491             ewtabD           = _fjsp_setzero_v2r8();
 492             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 493             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 494             ewtabFn          = _fjsp_setzero_v2r8();
 495             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 496             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 497             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 498             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(rinv10,velec));
 499             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 500
 501             /* Update potential sum for this i atom from the interaction with this j atom. */
 502             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 503             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 504
 505             fscal            = felec;
 506
 507             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 508
 509             /* Update vectorial force */
 510             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 511             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 512             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 513
 514             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 515             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 516             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 517
 518             /**************************
 519              * CALCULATE INTERACTIONS *
 520              **************************/
 521
 522             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 523
 524             /* Compute parameters for interactions between i and j atoms */
 525             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 526
 527             /* EWALD ELECTROSTATICS */
 528
 529             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 530             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 531             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 532             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 533             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 534
 535             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 536             ewtabD           = _fjsp_setzero_v2r8();
 537             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 538             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 539             ewtabFn          = _fjsp_setzero_v2r8();
 540             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 541             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 542             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 543             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(rinv20,velec));
 544             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 545
 546             /* Update potential sum for this i atom from the interaction with this j atom. */
 547             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 548             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 549
 550             fscal            = felec;
 551
 552             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 553
 554             /* Update vectorial force */
 555             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 556             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 557             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 558
 559             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 560             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 561             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 562
 563             /**************************
 564              * CALCULATE INTERACTIONS *
 565              **************************/
 566
 567             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 568
 569             /* Compute parameters for interactions between i and j atoms */
 570             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 571
 572             /* EWALD ELECTROSTATICS */
 573
 574             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 575             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 576             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 577             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 578             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 579
 580             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 581             ewtabD           = _fjsp_setzero_v2r8();
 582             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 583             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 584             ewtabFn          = _fjsp_setzero_v2r8();
 585             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 586             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 587             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 588             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(rinv30,velec));
 589             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 590
 591             /* Update potential sum for this i atom from the interaction with this j atom. */
 592             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 593             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 594
 595             fscal            = felec;
 596
 597             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 598
 599             /* Update vectorial force */
 600             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 601             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 602             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 603
 604             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 605             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 606             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 607
 608             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 609
 610             /* Inner loop uses 170 flops */
 611         }
 612
 613         /* End of innermost loop */
 614
 615         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 616                                               f+i_coord_offset,fshift+i_shift_offset);
 617
 618         ggid                        = gid[iidx];
 619         /* Update potential energies */
 620         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 621         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 622
 623         /* Increment number of inner iterations */
 624         inneriter                  += j_index_end - j_index_start;
 625
 626         /* Outer loop uses 26 flops */
 627     }
 628
 629     /* Increment number of outer iterations */
 630     outeriter        += nri;
 631
 632     /* Update outer/inner flops */
 633
 634     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*170);
 635 }
 636 /*
 637  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 638  * Electrostatics interaction: Ewald
 639  * VdW interaction:            LennardJones
 640  * Geometry:                   Water4-Particle
 641  * Calculate force/pot:        Force
 642  */
 643 void
 644 nb_kernel_ElecEw_VdwLJ_GeomW4P1_F_sparc64_hpc_ace_double
 645                     (t_nblist                    * gmx_restrict       nlist,
 646                      rvec                        * gmx_restrict          xx,
 647                      rvec                        * gmx_restrict          ff,
 648                      t_forcerec                  * gmx_restrict          fr,
 649                      t_mdatoms                   * gmx_restrict     mdatoms,
 650                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 651                      t_nrnb                      * gmx_restrict        nrnb)
 652 {
 653     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 654      * just 0 for non-waters.
 655      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 656      * jnr indices corresponding to data put in the four positions in the SIMD register.
 657      */
 658     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 659     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 660     int              jnrA,jnrB;
 661     int              j_coord_offsetA,j_coord_offsetB;
 662     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 663     real             rcutoff_scalar;
 664     real             *shiftvec,*fshift,*x,*f;
 665     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 666     int              vdwioffset0;
 667     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 668     int              vdwioffset1;
 669     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 670     int              vdwioffset2;
 671     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 672     int              vdwioffset3;
 673     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 674     int              vdwjidx0A,vdwjidx0B;
 675     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 676     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 677     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 678     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 679     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 680     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 681     real             *charge;
 682     int              nvdwtype;
 683     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 684     int              *vdwtype;
 685     real             *vdwparam;
 686     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 687     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 688     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 689     real             *ewtab;
 690     _fjsp_v2r8       itab_tmp;
 691     _fjsp_v2r8       dummy_mask,cutoff_mask;
 692     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 693     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 694     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 695
 696     x                = xx[0];
 697     f                = ff[0];
 698
 699     nri              = nlist->nri;
 700     iinr             = nlist->iinr;
 701     jindex           = nlist->jindex;
 702     jjnr             = nlist->jjnr;
 703     shiftidx         = nlist->shift;
 704     gid              = nlist->gid;
 705     shiftvec         = fr->shift_vec[0];
 706     fshift           = fr->fshift[0];
 707     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 708     charge           = mdatoms->chargeA;
 709     nvdwtype         = fr->ntype;
 710     vdwparam         = fr->nbfp;
 711     vdwtype          = mdatoms->typeA;
 712
 713     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 714     ewtab            = fr->ic->tabq_coul_F;
 715     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 716     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 717
 718     /* Setup water-specific parameters */
 719     inr              = nlist->iinr[0];
 720     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 721     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 722     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 723     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 724
 725     /* Avoid stupid compiler warnings */
 726     jnrA = jnrB = 0;
 727     j_coord_offsetA = 0;
 728     j_coord_offsetB = 0;
 729
 730     outeriter        = 0;
 731     inneriter        = 0;
 732
 733     /* Start outer loop over neighborlists */
 734     for(iidx=0; iidx<nri; iidx++)
 735     {
 736         /* Load shift vector for this list */
 737         i_shift_offset   = DIM*shiftidx[iidx];
 738
 739         /* Load limits for loop over neighbors */
 740         j_index_start    = jindex[iidx];
 741         j_index_end      = jindex[iidx+1];
 742
 743         /* Get outer coordinate index */
 744         inr              = iinr[iidx];
 745         i_coord_offset   = DIM*inr;
 746
 747         /* Load i particle coords and add shift vector */
 748         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 749                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 750
 751         fix0             = _fjsp_setzero_v2r8();
 752         fiy0             = _fjsp_setzero_v2r8();
 753         fiz0             = _fjsp_setzero_v2r8();
 754         fix1             = _fjsp_setzero_v2r8();
 755         fiy1             = _fjsp_setzero_v2r8();
 756         fiz1             = _fjsp_setzero_v2r8();
 757         fix2             = _fjsp_setzero_v2r8();
 758         fiy2             = _fjsp_setzero_v2r8();
 759         fiz2             = _fjsp_setzero_v2r8();
 760         fix3             = _fjsp_setzero_v2r8();
 761         fiy3             = _fjsp_setzero_v2r8();
 762         fiz3             = _fjsp_setzero_v2r8();
 763
 764         /* Start inner kernel loop */
 765         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 766         {
 767
 768             /* Get j neighbor index, and coordinate index */
 769             jnrA             = jjnr[jidx];
 770             jnrB             = jjnr[jidx+1];
 771             j_coord_offsetA  = DIM*jnrA;
 772             j_coord_offsetB  = DIM*jnrB;
 773
 774             /* load j atom coordinates */
 775             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 776                                               &jx0,&jy0,&jz0);
 777
 778             /* Calculate displacement vector */
 779             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 780             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 781             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 782             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 783             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 784             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 785             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 786             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 787             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 788             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 789             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 790             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 791
 792             /* Calculate squared distance and things based on it */
 793             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 794             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 795             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 796             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 797
 798             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 799             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 800             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 801
 802             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 803             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 804             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 805             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 806
 807             /* Load parameters for j particles */
 808             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 809             vdwjidx0A        = 2*vdwtype[jnrA+0];
 810             vdwjidx0B        = 2*vdwtype[jnrB+0];
 811
 812             fjx0             = _fjsp_setzero_v2r8();
 813             fjy0             = _fjsp_setzero_v2r8();
 814             fjz0             = _fjsp_setzero_v2r8();
 815
 816             /**************************
 817              * CALCULATE INTERACTIONS *
 818              **************************/
 819
 820             /* Compute parameters for interactions between i and j atoms */
 821             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 822                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 823
 824             /* LENNARD-JONES DISPERSION/REPULSION */
 825
 826             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 827             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
 828
 829             fscal            = fvdw;
 830
 831             /* Update vectorial force */
 832             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 833             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 834             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 835
 836             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 837             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 838             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 839
 840             /**************************
 841              * CALCULATE INTERACTIONS *
 842              **************************/
 843
 844             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
 845
 846             /* Compute parameters for interactions between i and j atoms */
 847             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 848
 849             /* EWALD ELECTROSTATICS */
 850
 851             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 852             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
 853             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 854             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 855             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 856
 857             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 858                                          &ewtabF,&ewtabFn);
 859             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 860             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
 861
 862             fscal            = felec;
 863
 864             /* Update vectorial force */
 865             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 866             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 867             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 868
 869             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 870             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 871             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 872
 873             /**************************
 874              * CALCULATE INTERACTIONS *
 875              **************************/
 876
 877             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
 878
 879             /* Compute parameters for interactions between i and j atoms */
 880             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 881
 882             /* EWALD ELECTROSTATICS */
 883
 884             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 885             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
 886             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 887             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 888             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 889
 890             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 891                                          &ewtabF,&ewtabFn);
 892             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 893             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
 894
 895             fscal            = felec;
 896
 897             /* Update vectorial force */
 898             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 899             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 900             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 901
 902             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 903             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 904             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 905
 906             /**************************
 907              * CALCULATE INTERACTIONS *
 908              **************************/
 909
 910             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
 911
 912             /* Compute parameters for interactions between i and j atoms */
 913             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 914
 915             /* EWALD ELECTROSTATICS */
 916
 917             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 918             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
 919             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 920             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 921             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 922
 923             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 924                                          &ewtabF,&ewtabFn);
 925             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 926             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
 927
 928             fscal            = felec;
 929
 930             /* Update vectorial force */
 931             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 932             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 933             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 934
 935             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 936             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 937             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 938
 939             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 940
 941             /* Inner loop uses 150 flops */
 942         }
 943
 944         if(jidx<j_index_end)
 945         {
 946
 947             jnrA             = jjnr[jidx];
 948             j_coord_offsetA  = DIM*jnrA;
 949
 950             /* load j atom coordinates */
 951             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 952                                               &jx0,&jy0,&jz0);
 953
 954             /* Calculate displacement vector */
 955             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 956             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 957             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 958             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 959             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 960             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 961             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 962             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 963             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 964             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 965             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 966             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 967
 968             /* Calculate squared distance and things based on it */
 969             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 970             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 971             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 972             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 973
 974             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 975             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 976             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 977
 978             rinvsq00         = gmx_fjsp_inv_v2r8(rsq00);
 979             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 980             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 981             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 982
 983             /* Load parameters for j particles */
 984             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 985             vdwjidx0A        = 2*vdwtype[jnrA+0];
 986
 987             fjx0             = _fjsp_setzero_v2r8();
 988             fjy0             = _fjsp_setzero_v2r8();
 989             fjz0             = _fjsp_setzero_v2r8();
 990
 991             /**************************
 992              * CALCULATE INTERACTIONS *
 993              **************************/
 994
 995             /* Compute parameters for interactions between i and j atoms */
 996             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 997
 998             /* LENNARD-JONES DISPERSION/REPULSION */
 999
1000             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
1001             fvdw             = _fjsp_mul_v2r8(_fjsp_msub_v2r8(c12_00,rinvsix,c6_00),_fjsp_mul_v2r8(rinvsix,rinvsq00));
1002
1003             fscal            = fvdw;
1004
1005             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1006
1007             /* Update vectorial force */
1008             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1009             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1010             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1011
1012             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1013             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1014             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1015
1016             /**************************
1017              * CALCULATE INTERACTIONS *
1018              **************************/
1019
1020             r10              = _fjsp_mul_v2r8(rsq10,rinv10);
1021
1022             /* Compute parameters for interactions between i and j atoms */
1023             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1024
1025             /* EWALD ELECTROSTATICS */
1026
1027             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1028             ewrt             = _fjsp_mul_v2r8(r10,ewtabscale);
1029             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1030             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1031             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1032
1033             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1034             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1035             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq10,rinv10),_fjsp_sub_v2r8(rinvsq10,felec));
1036
1037             fscal            = felec;
1038
1039             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1040
1041             /* Update vectorial force */
1042             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1043             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1044             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1045
1046             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1047             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1048             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1049
1050             /**************************
1051              * CALCULATE INTERACTIONS *
1052              **************************/
1053
1054             r20              = _fjsp_mul_v2r8(rsq20,rinv20);
1055
1056             /* Compute parameters for interactions between i and j atoms */
1057             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1058
1059             /* EWALD ELECTROSTATICS */
1060
1061             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1062             ewrt             = _fjsp_mul_v2r8(r20,ewtabscale);
1063             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1064             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1065             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1066
1067             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1068             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1069             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq20,rinv20),_fjsp_sub_v2r8(rinvsq20,felec));
1070
1071             fscal            = felec;
1072
1073             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1074
1075             /* Update vectorial force */
1076             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1077             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1078             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1079
1080             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1081             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1082             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1083
1084             /**************************
1085              * CALCULATE INTERACTIONS *
1086              **************************/
1087
1088             r30              = _fjsp_mul_v2r8(rsq30,rinv30);
1089
1090             /* Compute parameters for interactions between i and j atoms */
1091             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1092
1093             /* EWALD ELECTROSTATICS */
1094
1095             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
1096             ewrt             = _fjsp_mul_v2r8(r30,ewtabscale);
1097             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
1098             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
1099             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
1100
1101             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
1102             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
1103             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq30,rinv30),_fjsp_sub_v2r8(rinvsq30,felec));
1104
1105             fscal            = felec;
1106
1107             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1108
1109             /* Update vectorial force */
1110             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1111             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1112             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1113
1114             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1115             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1116             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1117
1118             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1119
1120             /* Inner loop uses 150 flops */
1121         }
1122
1123         /* End of innermost loop */
1124
1125         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1126                                               f+i_coord_offset,fshift+i_shift_offset);
1127
1128         /* Increment number of inner iterations */
1129         inneriter                  += j_index_end - j_index_start;
1130
1131         /* Outer loop uses 24 flops */
1132     }
1133
1134     /* Increment number of outer iterations */
1135     outeriter        += nri;
1136
1137     /* Update outer/inner flops */
1138
1139     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*150);
1140 }