src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwCSTab_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 "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_ElecRFCut_VdwCSTab_GeomW4P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            CubicSplineTable
  53  * Geometry:                   Water4-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwCSTab_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       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 102     real             *vftab;
 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     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 123     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 124     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 125     nvdwtype         = fr->ntype;
 126     vdwparam         = fr->nbfp;
 127     vdwtype          = mdatoms->typeA;
 128
 129     vftab            = kernel_data->table_vdw->data;
 130     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 131
 132     /* Setup water-specific parameters */
 133     inr              = nlist->iinr[0];
 134     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 135     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 136     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 137     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 138
 139     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 140     rcutoff_scalar   = fr->rcoulomb;
 141     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 142     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 143
 144     /* Avoid stupid compiler warnings */
 145     jnrA = jnrB = 0;
 146     j_coord_offsetA = 0;
 147     j_coord_offsetB = 0;
 148
 149     outeriter        = 0;
 150     inneriter        = 0;
 151
 152     /* Start outer loop over neighborlists */
 153     for(iidx=0; iidx<nri; iidx++)
 154     {
 155         /* Load shift vector for this list */
 156         i_shift_offset   = DIM*shiftidx[iidx];
 157
 158         /* Load limits for loop over neighbors */
 159         j_index_start    = jindex[iidx];
 160         j_index_end      = jindex[iidx+1];
 161
 162         /* Get outer coordinate index */
 163         inr              = iinr[iidx];
 164         i_coord_offset   = DIM*inr;
 165
 166         /* Load i particle coords and add shift vector */
 167         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 168                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 169
 170         fix0             = _fjsp_setzero_v2r8();
 171         fiy0             = _fjsp_setzero_v2r8();
 172         fiz0             = _fjsp_setzero_v2r8();
 173         fix1             = _fjsp_setzero_v2r8();
 174         fiy1             = _fjsp_setzero_v2r8();
 175         fiz1             = _fjsp_setzero_v2r8();
 176         fix2             = _fjsp_setzero_v2r8();
 177         fiy2             = _fjsp_setzero_v2r8();
 178         fiz2             = _fjsp_setzero_v2r8();
 179         fix3             = _fjsp_setzero_v2r8();
 180         fiy3             = _fjsp_setzero_v2r8();
 181         fiz3             = _fjsp_setzero_v2r8();
 182
 183         /* Reset potential sums */
 184         velecsum         = _fjsp_setzero_v2r8();
 185         vvdwsum          = _fjsp_setzero_v2r8();
 186
 187         /* Start inner kernel loop */
 188         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 189         {
 190
 191             /* Get j neighbor index, and coordinate index */
 192             jnrA             = jjnr[jidx];
 193             jnrB             = jjnr[jidx+1];
 194             j_coord_offsetA  = DIM*jnrA;
 195             j_coord_offsetB  = DIM*jnrB;
 196
 197             /* load j atom coordinates */
 198             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 199                                               &jx0,&jy0,&jz0);
 200
 201             /* Calculate displacement vector */
 202             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 203             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 204             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 205             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 206             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 207             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 208             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 209             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 210             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 211             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 212             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 213             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 214
 215             /* Calculate squared distance and things based on it */
 216             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 217             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 218             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 219             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 220
 221             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 222             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 223             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 224             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 225
 226             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 227             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 228             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 229
 230             /* Load parameters for j particles */
 231             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 232             vdwjidx0A        = 2*vdwtype[jnrA+0];
 233             vdwjidx0B        = 2*vdwtype[jnrB+0];
 234
 235             fjx0             = _fjsp_setzero_v2r8();
 236             fjy0             = _fjsp_setzero_v2r8();
 237             fjz0             = _fjsp_setzero_v2r8();
 238
 239             /**************************
 240              * CALCULATE INTERACTIONS *
 241              **************************/
 242
 243             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 244
 245             /* Compute parameters for interactions between i and j atoms */
 246             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 247                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 248
 249             /* Calculate table index by multiplying r with table scale and truncate to integer */
 250             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 251             itab_tmp         = _fjsp_dtox_v2r8(rt);
 252             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 253             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 254             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 255
 256             vfconv.i[0]     *= 8;
 257             vfconv.i[1]     *= 8;
 258
 259             /* CUBIC SPLINE TABLE DISPERSION */
 260             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 261             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 262             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 263             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 264             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 265             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 266             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 267             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 268             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 269             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 270             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 271
 272             /* CUBIC SPLINE TABLE REPULSION */
 273             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 274             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 275             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 276             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 277             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 278             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 279             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 280             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 281             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 282             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 283             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 284             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 285             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 286
 287             /* Update potential sum for this i atom from the interaction with this j atom. */
 288             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 289
 290             fscal            = fvdw;
 291
 292             /* Update vectorial force */
 293             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 294             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 295             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 296
 297             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 298             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 299             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 300
 301             /**************************
 302              * CALCULATE INTERACTIONS *
 303              **************************/
 304
 305             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 306             {
 307
 308             /* Compute parameters for interactions between i and j atoms */
 309             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 310
 311             /* REACTION-FIELD ELECTROSTATICS */
 312             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 313             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 314
 315             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 316
 317             /* Update potential sum for this i atom from the interaction with this j atom. */
 318             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 319             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 320
 321             fscal            = felec;
 322
 323             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 324
 325             /* Update vectorial force */
 326             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 327             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 328             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 329
 330             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 331             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 332             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 333
 334             }
 335
 336             /**************************
 337              * CALCULATE INTERACTIONS *
 338              **************************/
 339
 340             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 341             {
 342
 343             /* Compute parameters for interactions between i and j atoms */
 344             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 345
 346             /* REACTION-FIELD ELECTROSTATICS */
 347             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 348             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 349
 350             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 351
 352             /* Update potential sum for this i atom from the interaction with this j atom. */
 353             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 354             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 355
 356             fscal            = felec;
 357
 358             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 359
 360             /* Update vectorial force */
 361             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 362             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 363             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 364
 365             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 366             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 367             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 368
 369             }
 370
 371             /**************************
 372              * CALCULATE INTERACTIONS *
 373              **************************/
 374
 375             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 376             {
 377
 378             /* Compute parameters for interactions between i and j atoms */
 379             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 380
 381             /* REACTION-FIELD ELECTROSTATICS */
 382             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 383             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 384
 385             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 386
 387             /* Update potential sum for this i atom from the interaction with this j atom. */
 388             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 389             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 390
 391             fscal            = felec;
 392
 393             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 394
 395             /* Update vectorial force */
 396             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 397             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 398             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 399
 400             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 401             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 402             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 403
 404             }
 405
 406             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 407
 408             /* Inner loop uses 179 flops */
 409         }
 410
 411         if(jidx<j_index_end)
 412         {
 413
 414             jnrA             = jjnr[jidx];
 415             j_coord_offsetA  = DIM*jnrA;
 416
 417             /* load j atom coordinates */
 418             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 419                                               &jx0,&jy0,&jz0);
 420
 421             /* Calculate displacement vector */
 422             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 423             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 424             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 425             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 426             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 427             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 428             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 429             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 430             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 431             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 432             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 433             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 434
 435             /* Calculate squared distance and things based on it */
 436             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 437             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 438             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 439             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 440
 441             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 442             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 443             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 444             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 445
 446             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 447             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 448             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 449
 450             /* Load parameters for j particles */
 451             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 452             vdwjidx0A        = 2*vdwtype[jnrA+0];
 453
 454             fjx0             = _fjsp_setzero_v2r8();
 455             fjy0             = _fjsp_setzero_v2r8();
 456             fjz0             = _fjsp_setzero_v2r8();
 457
 458             /**************************
 459              * CALCULATE INTERACTIONS *
 460              **************************/
 461
 462             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 463
 464             /* Compute parameters for interactions between i and j atoms */
 465             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 466
 467             /* Calculate table index by multiplying r with table scale and truncate to integer */
 468             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 469             itab_tmp         = _fjsp_dtox_v2r8(rt);
 470             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 471             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 472             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 473
 474             vfconv.i[0]     *= 8;
 475             vfconv.i[1]     *= 8;
 476
 477             /* CUBIC SPLINE TABLE DISPERSION */
 478             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 479             F                = _fjsp_setzero_v2r8();
 480             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 481             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 482             H                = _fjsp_setzero_v2r8();
 483             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 484             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 485             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 486             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 487             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 488             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 489
 490             /* CUBIC SPLINE TABLE REPULSION */
 491             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 492             F                = _fjsp_setzero_v2r8();
 493             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 494             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 495             H                = _fjsp_setzero_v2r8();
 496             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 497             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 498             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 499             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 500             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 501             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 502             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 503             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 504
 505             /* Update potential sum for this i atom from the interaction with this j atom. */
 506             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 507             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 508
 509             fscal            = fvdw;
 510
 511             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 512
 513             /* Update vectorial force */
 514             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 515             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 516             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 517
 518             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 519             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 520             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 521
 522             /**************************
 523              * CALCULATE INTERACTIONS *
 524              **************************/
 525
 526             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 527             {
 528
 529             /* Compute parameters for interactions between i and j atoms */
 530             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 531
 532             /* REACTION-FIELD ELECTROSTATICS */
 533             velec            = _fjsp_mul_v2r8(qq10,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq10,rinv10),crf));
 534             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 535
 536             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 537
 538             /* Update potential sum for this i atom from the interaction with this j atom. */
 539             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 540             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 541             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 542
 543             fscal            = felec;
 544
 545             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 546
 547             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 548
 549             /* Update vectorial force */
 550             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 551             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 552             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 553
 554             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 555             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 556             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 557
 558             }
 559
 560             /**************************
 561              * CALCULATE INTERACTIONS *
 562              **************************/
 563
 564             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 565             {
 566
 567             /* Compute parameters for interactions between i and j atoms */
 568             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 569
 570             /* REACTION-FIELD ELECTROSTATICS */
 571             velec            = _fjsp_mul_v2r8(qq20,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq20,rinv20),crf));
 572             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 573
 574             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 575
 576             /* Update potential sum for this i atom from the interaction with this j atom. */
 577             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 578             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 579             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 580
 581             fscal            = felec;
 582
 583             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 584
 585             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 586
 587             /* Update vectorial force */
 588             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 589             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 590             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 591
 592             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 593             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 594             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 595
 596             }
 597
 598             /**************************
 599              * CALCULATE INTERACTIONS *
 600              **************************/
 601
 602             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 603             {
 604
 605             /* Compute parameters for interactions between i and j atoms */
 606             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 607
 608             /* REACTION-FIELD ELECTROSTATICS */
 609             velec            = _fjsp_mul_v2r8(qq30,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq30,rinv30),crf));
 610             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 611
 612             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 613
 614             /* Update potential sum for this i atom from the interaction with this j atom. */
 615             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 616             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 617             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 618
 619             fscal            = felec;
 620
 621             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 622
 623             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 624
 625             /* Update vectorial force */
 626             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 627             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 628             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 629
 630             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 631             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 632             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 633
 634             }
 635
 636             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
 637
 638             /* Inner loop uses 179 flops */
 639         }
 640
 641         /* End of innermost loop */
 642
 643         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
 644                                               f+i_coord_offset,fshift+i_shift_offset);
 645
 646         ggid                        = gid[iidx];
 647         /* Update potential energies */
 648         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 649         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 650
 651         /* Increment number of inner iterations */
 652         inneriter                  += j_index_end - j_index_start;
 653
 654         /* Outer loop uses 26 flops */
 655     }
 656
 657     /* Increment number of outer iterations */
 658     outeriter        += nri;
 659
 660     /* Update outer/inner flops */
 661
 662     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_VF,outeriter*26 + inneriter*179);
 663 }
 664 /*
 665  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 666  * Electrostatics interaction: ReactionField
 667  * VdW interaction:            CubicSplineTable
 668  * Geometry:                   Water4-Particle
 669  * Calculate force/pot:        Force
 670  */
 671 void
 672 nb_kernel_ElecRFCut_VdwCSTab_GeomW4P1_F_sparc64_hpc_ace_double
 673                     (t_nblist                    * gmx_restrict       nlist,
 674                      rvec                        * gmx_restrict          xx,
 675                      rvec                        * gmx_restrict          ff,
 676                      t_forcerec                  * gmx_restrict          fr,
 677                      t_mdatoms                   * gmx_restrict     mdatoms,
 678                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 679                      t_nrnb                      * gmx_restrict        nrnb)
 680 {
 681     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 682      * just 0 for non-waters.
 683      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 684      * jnr indices corresponding to data put in the four positions in the SIMD register.
 685      */
 686     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 687     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 688     int              jnrA,jnrB;
 689     int              j_coord_offsetA,j_coord_offsetB;
 690     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 691     real             rcutoff_scalar;
 692     real             *shiftvec,*fshift,*x,*f;
 693     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 694     int              vdwioffset0;
 695     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 696     int              vdwioffset1;
 697     _fjsp_v2r8       ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
 698     int              vdwioffset2;
 699     _fjsp_v2r8       ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
 700     int              vdwioffset3;
 701     _fjsp_v2r8       ix3,iy3,iz3,fix3,fiy3,fiz3,iq3,isai3;
 702     int              vdwjidx0A,vdwjidx0B;
 703     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 704     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 705     _fjsp_v2r8       dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
 706     _fjsp_v2r8       dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
 707     _fjsp_v2r8       dx30,dy30,dz30,rsq30,rinv30,rinvsq30,r30,qq30,c6_30,c12_30;
 708     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 709     real             *charge;
 710     int              nvdwtype;
 711     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 712     int              *vdwtype;
 713     real             *vdwparam;
 714     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 715     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 716     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 717     real             *vftab;
 718     _fjsp_v2r8       itab_tmp;
 719     _fjsp_v2r8       dummy_mask,cutoff_mask;
 720     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 721     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 722     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 723
 724     x                = xx[0];
 725     f                = ff[0];
 726
 727     nri              = nlist->nri;
 728     iinr             = nlist->iinr;
 729     jindex           = nlist->jindex;
 730     jjnr             = nlist->jjnr;
 731     shiftidx         = nlist->shift;
 732     gid              = nlist->gid;
 733     shiftvec         = fr->shift_vec[0];
 734     fshift           = fr->fshift[0];
 735     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 736     charge           = mdatoms->chargeA;
 737     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 738     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 739     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 740     nvdwtype         = fr->ntype;
 741     vdwparam         = fr->nbfp;
 742     vdwtype          = mdatoms->typeA;
 743
 744     vftab            = kernel_data->table_vdw->data;
 745     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 746
 747     /* Setup water-specific parameters */
 748     inr              = nlist->iinr[0];
 749     iq1              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+1]));
 750     iq2              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+2]));
 751     iq3              = _fjsp_mul_v2r8(facel,gmx_fjsp_set1_v2r8(charge[inr+3]));
 752     vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 753
 754     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 755     rcutoff_scalar   = fr->rcoulomb;
 756     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 757     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 758
 759     /* Avoid stupid compiler warnings */
 760     jnrA = jnrB = 0;
 761     j_coord_offsetA = 0;
 762     j_coord_offsetB = 0;
 763
 764     outeriter        = 0;
 765     inneriter        = 0;
 766
 767     /* Start outer loop over neighborlists */
 768     for(iidx=0; iidx<nri; iidx++)
 769     {
 770         /* Load shift vector for this list */
 771         i_shift_offset   = DIM*shiftidx[iidx];
 772
 773         /* Load limits for loop over neighbors */
 774         j_index_start    = jindex[iidx];
 775         j_index_end      = jindex[iidx+1];
 776
 777         /* Get outer coordinate index */
 778         inr              = iinr[iidx];
 779         i_coord_offset   = DIM*inr;
 780
 781         /* Load i particle coords and add shift vector */
 782         gmx_fjsp_load_shift_and_4rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,
 783                                                  &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2,&ix3,&iy3,&iz3);
 784
 785         fix0             = _fjsp_setzero_v2r8();
 786         fiy0             = _fjsp_setzero_v2r8();
 787         fiz0             = _fjsp_setzero_v2r8();
 788         fix1             = _fjsp_setzero_v2r8();
 789         fiy1             = _fjsp_setzero_v2r8();
 790         fiz1             = _fjsp_setzero_v2r8();
 791         fix2             = _fjsp_setzero_v2r8();
 792         fiy2             = _fjsp_setzero_v2r8();
 793         fiz2             = _fjsp_setzero_v2r8();
 794         fix3             = _fjsp_setzero_v2r8();
 795         fiy3             = _fjsp_setzero_v2r8();
 796         fiz3             = _fjsp_setzero_v2r8();
 797
 798         /* Start inner kernel loop */
 799         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 800         {
 801
 802             /* Get j neighbor index, and coordinate index */
 803             jnrA             = jjnr[jidx];
 804             jnrB             = jjnr[jidx+1];
 805             j_coord_offsetA  = DIM*jnrA;
 806             j_coord_offsetB  = DIM*jnrB;
 807
 808             /* load j atom coordinates */
 809             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 810                                               &jx0,&jy0,&jz0);
 811
 812             /* Calculate displacement vector */
 813             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 814             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 815             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 816             dx10             = _fjsp_sub_v2r8(ix1,jx0);
 817             dy10             = _fjsp_sub_v2r8(iy1,jy0);
 818             dz10             = _fjsp_sub_v2r8(iz1,jz0);
 819             dx20             = _fjsp_sub_v2r8(ix2,jx0);
 820             dy20             = _fjsp_sub_v2r8(iy2,jy0);
 821             dz20             = _fjsp_sub_v2r8(iz2,jz0);
 822             dx30             = _fjsp_sub_v2r8(ix3,jx0);
 823             dy30             = _fjsp_sub_v2r8(iy3,jy0);
 824             dz30             = _fjsp_sub_v2r8(iz3,jz0);
 825
 826             /* Calculate squared distance and things based on it */
 827             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 828             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
 829             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
 830             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
 831
 832             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 833             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
 834             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
 835             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
 836
 837             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
 838             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
 839             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
 840
 841             /* Load parameters for j particles */
 842             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 843             vdwjidx0A        = 2*vdwtype[jnrA+0];
 844             vdwjidx0B        = 2*vdwtype[jnrB+0];
 845
 846             fjx0             = _fjsp_setzero_v2r8();
 847             fjy0             = _fjsp_setzero_v2r8();
 848             fjz0             = _fjsp_setzero_v2r8();
 849
 850             /**************************
 851              * CALCULATE INTERACTIONS *
 852              **************************/
 853
 854             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 855
 856             /* Compute parameters for interactions between i and j atoms */
 857             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 858                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 859
 860             /* Calculate table index by multiplying r with table scale and truncate to integer */
 861             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 862             itab_tmp         = _fjsp_dtox_v2r8(rt);
 863             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 864             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 865             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 866
 867             vfconv.i[0]     *= 8;
 868             vfconv.i[1]     *= 8;
 869
 870             /* CUBIC SPLINE TABLE DISPERSION */
 871             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 872             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 873             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 874             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 875             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 876             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 877             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 878             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 879             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 880
 881             /* CUBIC SPLINE TABLE REPULSION */
 882             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 883             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 884             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 885             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 886             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 887             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 888             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 889             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 890             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 891             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 892
 893             fscal            = fvdw;
 894
 895             /* Update vectorial force */
 896             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 897             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 898             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 899
 900             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
 901             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
 902             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
 903
 904             /**************************
 905              * CALCULATE INTERACTIONS *
 906              **************************/
 907
 908             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
 909             {
 910
 911             /* Compute parameters for interactions between i and j atoms */
 912             qq10             = _fjsp_mul_v2r8(iq1,jq0);
 913
 914             /* REACTION-FIELD ELECTROSTATICS */
 915             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
 916
 917             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
 918
 919             fscal            = felec;
 920
 921             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 922
 923             /* Update vectorial force */
 924             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
 925             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
 926             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
 927
 928             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
 929             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
 930             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
 931
 932             }
 933
 934             /**************************
 935              * CALCULATE INTERACTIONS *
 936              **************************/
 937
 938             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
 939             {
 940
 941             /* Compute parameters for interactions between i and j atoms */
 942             qq20             = _fjsp_mul_v2r8(iq2,jq0);
 943
 944             /* REACTION-FIELD ELECTROSTATICS */
 945             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
 946
 947             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
 948
 949             fscal            = felec;
 950
 951             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 952
 953             /* Update vectorial force */
 954             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
 955             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
 956             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
 957
 958             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
 959             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
 960             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
 961
 962             }
 963
 964             /**************************
 965              * CALCULATE INTERACTIONS *
 966              **************************/
 967
 968             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
 969             {
 970
 971             /* Compute parameters for interactions between i and j atoms */
 972             qq30             = _fjsp_mul_v2r8(iq3,jq0);
 973
 974             /* REACTION-FIELD ELECTROSTATICS */
 975             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
 976
 977             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
 978
 979             fscal            = felec;
 980
 981             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 982
 983             /* Update vectorial force */
 984             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
 985             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
 986             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
 987
 988             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
 989             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
 990             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
 991
 992             }
 993
 994             gmx_fjsp_decrement_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
 995
 996             /* Inner loop uses 153 flops */
 997         }
 998
 999         if(jidx<j_index_end)
1000         {
1001
1002             jnrA             = jjnr[jidx];
1003             j_coord_offsetA  = DIM*jnrA;
1004
1005             /* load j atom coordinates */
1006             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
1007                                               &jx0,&jy0,&jz0);
1008
1009             /* Calculate displacement vector */
1010             dx00             = _fjsp_sub_v2r8(ix0,jx0);
1011             dy00             = _fjsp_sub_v2r8(iy0,jy0);
1012             dz00             = _fjsp_sub_v2r8(iz0,jz0);
1013             dx10             = _fjsp_sub_v2r8(ix1,jx0);
1014             dy10             = _fjsp_sub_v2r8(iy1,jy0);
1015             dz10             = _fjsp_sub_v2r8(iz1,jz0);
1016             dx20             = _fjsp_sub_v2r8(ix2,jx0);
1017             dy20             = _fjsp_sub_v2r8(iy2,jy0);
1018             dz20             = _fjsp_sub_v2r8(iz2,jz0);
1019             dx30             = _fjsp_sub_v2r8(ix3,jx0);
1020             dy30             = _fjsp_sub_v2r8(iy3,jy0);
1021             dz30             = _fjsp_sub_v2r8(iz3,jz0);
1022
1023             /* Calculate squared distance and things based on it */
1024             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
1025             rsq10            = gmx_fjsp_calc_rsq_v2r8(dx10,dy10,dz10);
1026             rsq20            = gmx_fjsp_calc_rsq_v2r8(dx20,dy20,dz20);
1027             rsq30            = gmx_fjsp_calc_rsq_v2r8(dx30,dy30,dz30);
1028
1029             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
1030             rinv10           = gmx_fjsp_invsqrt_v2r8(rsq10);
1031             rinv20           = gmx_fjsp_invsqrt_v2r8(rsq20);
1032             rinv30           = gmx_fjsp_invsqrt_v2r8(rsq30);
1033
1034             rinvsq10         = _fjsp_mul_v2r8(rinv10,rinv10);
1035             rinvsq20         = _fjsp_mul_v2r8(rinv20,rinv20);
1036             rinvsq30         = _fjsp_mul_v2r8(rinv30,rinv30);
1037
1038             /* Load parameters for j particles */
1039             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
1040             vdwjidx0A        = 2*vdwtype[jnrA+0];
1041
1042             fjx0             = _fjsp_setzero_v2r8();
1043             fjy0             = _fjsp_setzero_v2r8();
1044             fjz0             = _fjsp_setzero_v2r8();
1045
1046             /**************************
1047              * CALCULATE INTERACTIONS *
1048              **************************/
1049
1050             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
1051
1052             /* Compute parameters for interactions between i and j atoms */
1053             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
1054
1055             /* Calculate table index by multiplying r with table scale and truncate to integer */
1056             rt               = _fjsp_mul_v2r8(r00,vftabscale);
1057             itab_tmp         = _fjsp_dtox_v2r8(rt);
1058             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
1059             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
1060             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
1061
1062             vfconv.i[0]     *= 8;
1063             vfconv.i[1]     *= 8;
1064
1065             /* CUBIC SPLINE TABLE DISPERSION */
1066             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
1067             F                = _fjsp_setzero_v2r8();
1068             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1069             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
1070             H                = _fjsp_setzero_v2r8();
1071             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1072             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1073             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1074             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
1075
1076             /* CUBIC SPLINE TABLE REPULSION */
1077             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
1078             F                = _fjsp_setzero_v2r8();
1079             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
1080             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
1081             H                = _fjsp_setzero_v2r8();
1082             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
1083             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
1084             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
1085             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
1086             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
1087
1088             fscal            = fvdw;
1089
1090             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1091
1092             /* Update vectorial force */
1093             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
1094             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
1095             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
1096
1097             fjx0             = _fjsp_madd_v2r8(dx00,fscal,fjx0);
1098             fjy0             = _fjsp_madd_v2r8(dy00,fscal,fjy0);
1099             fjz0             = _fjsp_madd_v2r8(dz00,fscal,fjz0);
1100
1101             /**************************
1102              * CALCULATE INTERACTIONS *
1103              **************************/
1104
1105             if (gmx_fjsp_any_lt_v2r8(rsq10,rcutoff2))
1106             {
1107
1108             /* Compute parameters for interactions between i and j atoms */
1109             qq10             = _fjsp_mul_v2r8(iq1,jq0);
1110
1111             /* REACTION-FIELD ELECTROSTATICS */
1112             felec            = _fjsp_mul_v2r8(qq10,_fjsp_msub_v2r8(rinv10,rinvsq10,krf2));
1113
1114             cutoff_mask      = _fjsp_cmplt_v2r8(rsq10,rcutoff2);
1115
1116             fscal            = felec;
1117
1118             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1119
1120             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1121
1122             /* Update vectorial force */
1123             fix1             = _fjsp_madd_v2r8(dx10,fscal,fix1);
1124             fiy1             = _fjsp_madd_v2r8(dy10,fscal,fiy1);
1125             fiz1             = _fjsp_madd_v2r8(dz10,fscal,fiz1);
1126
1127             fjx0             = _fjsp_madd_v2r8(dx10,fscal,fjx0);
1128             fjy0             = _fjsp_madd_v2r8(dy10,fscal,fjy0);
1129             fjz0             = _fjsp_madd_v2r8(dz10,fscal,fjz0);
1130
1131             }
1132
1133             /**************************
1134              * CALCULATE INTERACTIONS *
1135              **************************/
1136
1137             if (gmx_fjsp_any_lt_v2r8(rsq20,rcutoff2))
1138             {
1139
1140             /* Compute parameters for interactions between i and j atoms */
1141             qq20             = _fjsp_mul_v2r8(iq2,jq0);
1142
1143             /* REACTION-FIELD ELECTROSTATICS */
1144             felec            = _fjsp_mul_v2r8(qq20,_fjsp_msub_v2r8(rinv20,rinvsq20,krf2));
1145
1146             cutoff_mask      = _fjsp_cmplt_v2r8(rsq20,rcutoff2);
1147
1148             fscal            = felec;
1149
1150             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1151
1152             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1153
1154             /* Update vectorial force */
1155             fix2             = _fjsp_madd_v2r8(dx20,fscal,fix2);
1156             fiy2             = _fjsp_madd_v2r8(dy20,fscal,fiy2);
1157             fiz2             = _fjsp_madd_v2r8(dz20,fscal,fiz2);
1158
1159             fjx0             = _fjsp_madd_v2r8(dx20,fscal,fjx0);
1160             fjy0             = _fjsp_madd_v2r8(dy20,fscal,fjy0);
1161             fjz0             = _fjsp_madd_v2r8(dz20,fscal,fjz0);
1162
1163             }
1164
1165             /**************************
1166              * CALCULATE INTERACTIONS *
1167              **************************/
1168
1169             if (gmx_fjsp_any_lt_v2r8(rsq30,rcutoff2))
1170             {
1171
1172             /* Compute parameters for interactions between i and j atoms */
1173             qq30             = _fjsp_mul_v2r8(iq3,jq0);
1174
1175             /* REACTION-FIELD ELECTROSTATICS */
1176             felec            = _fjsp_mul_v2r8(qq30,_fjsp_msub_v2r8(rinv30,rinvsq30,krf2));
1177
1178             cutoff_mask      = _fjsp_cmplt_v2r8(rsq30,rcutoff2);
1179
1180             fscal            = felec;
1181
1182             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
1183
1184             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
1185
1186             /* Update vectorial force */
1187             fix3             = _fjsp_madd_v2r8(dx30,fscal,fix3);
1188             fiy3             = _fjsp_madd_v2r8(dy30,fscal,fiy3);
1189             fiz3             = _fjsp_madd_v2r8(dz30,fscal,fiz3);
1190
1191             fjx0             = _fjsp_madd_v2r8(dx30,fscal,fjx0);
1192             fjy0             = _fjsp_madd_v2r8(dy30,fscal,fjy0);
1193             fjz0             = _fjsp_madd_v2r8(dz30,fscal,fjz0);
1194
1195             }
1196
1197             gmx_fjsp_decrement_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fjx0,fjy0,fjz0);
1198
1199             /* Inner loop uses 153 flops */
1200         }
1201
1202         /* End of innermost loop */
1203
1204         gmx_fjsp_update_iforce_4atom_swizzle_v2r8(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,fix3,fiy3,fiz3,
1205                                               f+i_coord_offset,fshift+i_shift_offset);
1206
1207         /* Increment number of inner iterations */
1208         inneriter                  += j_index_end - j_index_start;
1209
1210         /* Outer loop uses 24 flops */
1211     }
1212
1213     /* Increment number of outer iterations */
1214     outeriter        += nri;
1215
1216     /* Update outer/inner flops */
1217
1218     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W4_F,outeriter*24 + inneriter*153);
1219 }