src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_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_VdwLJSw_GeomP1P1_VF_sparc64_hpc_ace_double
  51  * Electrostatics interaction: ReactionField
  52  * VdW interaction:            LennardJones
  53  * Geometry:                   Particle-Particle
  54  * Calculate force/pot:        PotentialAndForce
  55  */
  56 void
  57 nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_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              vdwjidx0A,vdwjidx0B;
  82     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  83     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  84     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  85     real             *charge;
  86     int              nvdwtype;
  87     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  88     int              *vdwtype;
  89     real             *vdwparam;
  90     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  91     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  92     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  93     real             rswitch_scalar,d_scalar;
  94     _fjsp_v2r8       itab_tmp;
  95     _fjsp_v2r8       dummy_mask,cutoff_mask;
  96     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  97     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
  98     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
  99
 100     x                = xx[0];
 101     f                = ff[0];
 102
 103     nri              = nlist->nri;
 104     iinr             = nlist->iinr;
 105     jindex           = nlist->jindex;
 106     jjnr             = nlist->jjnr;
 107     shiftidx         = nlist->shift;
 108     gid              = nlist->gid;
 109     shiftvec         = fr->shift_vec[0];
 110     fshift           = fr->fshift[0];
 111     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 112     charge           = mdatoms->chargeA;
 113     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 114     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 115     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 116     nvdwtype         = fr->ntype;
 117     vdwparam         = fr->nbfp;
 118     vdwtype          = mdatoms->typeA;
 119
 120     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 121     rcutoff_scalar   = fr->rcoulomb;
 122     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 123     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 124
 125     rswitch_scalar   = fr->rvdw_switch;
 126     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 127     /* Setup switch parameters */
 128     d_scalar         = rcutoff_scalar-rswitch_scalar;
 129     d                = gmx_fjsp_set1_v2r8(d_scalar);
 130     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 131     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 132     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 133     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 134     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 135     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 136
 137     /* Avoid stupid compiler warnings */
 138     jnrA = jnrB = 0;
 139     j_coord_offsetA = 0;
 140     j_coord_offsetB = 0;
 141
 142     outeriter        = 0;
 143     inneriter        = 0;
 144
 145     /* Start outer loop over neighborlists */
 146     for(iidx=0; iidx<nri; iidx++)
 147     {
 148         /* Load shift vector for this list */
 149         i_shift_offset   = DIM*shiftidx[iidx];
 150
 151         /* Load limits for loop over neighbors */
 152         j_index_start    = jindex[iidx];
 153         j_index_end      = jindex[iidx+1];
 154
 155         /* Get outer coordinate index */
 156         inr              = iinr[iidx];
 157         i_coord_offset   = DIM*inr;
 158
 159         /* Load i particle coords and add shift vector */
 160         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 161
 162         fix0             = _fjsp_setzero_v2r8();
 163         fiy0             = _fjsp_setzero_v2r8();
 164         fiz0             = _fjsp_setzero_v2r8();
 165
 166         /* Load parameters for i particles */
 167         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 168         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 169
 170         /* Reset potential sums */
 171         velecsum         = _fjsp_setzero_v2r8();
 172         vvdwsum          = _fjsp_setzero_v2r8();
 173
 174         /* Start inner kernel loop */
 175         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 176         {
 177
 178             /* Get j neighbor index, and coordinate index */
 179             jnrA             = jjnr[jidx];
 180             jnrB             = jjnr[jidx+1];
 181             j_coord_offsetA  = DIM*jnrA;
 182             j_coord_offsetB  = DIM*jnrB;
 183
 184             /* load j atom coordinates */
 185             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 186                                               &jx0,&jy0,&jz0);
 187
 188             /* Calculate displacement vector */
 189             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 190             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 191             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 192
 193             /* Calculate squared distance and things based on it */
 194             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 195
 196             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 197
 198             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 199
 200             /* Load parameters for j particles */
 201             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 202             vdwjidx0A        = 2*vdwtype[jnrA+0];
 203             vdwjidx0B        = 2*vdwtype[jnrB+0];
 204
 205             /**************************
 206              * CALCULATE INTERACTIONS *
 207              **************************/
 208
 209             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 210             {
 211
 212             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 213
 214             /* Compute parameters for interactions between i and j atoms */
 215             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 216             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 217                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 218
 219             /* REACTION-FIELD ELECTROSTATICS */
 220             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 221             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 222
 223             /* LENNARD-JONES DISPERSION/REPULSION */
 224
 225             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 226             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 227             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 228             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 229             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 230
 231             d                = _fjsp_sub_v2r8(r00,rswitch);
 232             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 233             d2               = _fjsp_mul_v2r8(d,d);
 234             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 235
 236             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 237
 238             /* Evaluate switch function */
 239             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 240             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 241             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 242             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 243
 244             /* Update potential sum for this i atom from the interaction with this j atom. */
 245             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 246             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 247             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 248             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 249
 250             fscal            = _fjsp_add_v2r8(felec,fvdw);
 251
 252             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 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             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 260
 261             }
 262
 263             /* Inner loop uses 73 flops */
 264         }
 265
 266         if(jidx<j_index_end)
 267         {
 268
 269             jnrA             = jjnr[jidx];
 270             j_coord_offsetA  = DIM*jnrA;
 271
 272             /* load j atom coordinates */
 273             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 274                                               &jx0,&jy0,&jz0);
 275
 276             /* Calculate displacement vector */
 277             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 278             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 279             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 280
 281             /* Calculate squared distance and things based on it */
 282             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 283
 284             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 285
 286             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 287
 288             /* Load parameters for j particles */
 289             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 290             vdwjidx0A        = 2*vdwtype[jnrA+0];
 291
 292             /**************************
 293              * CALCULATE INTERACTIONS *
 294              **************************/
 295
 296             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 297             {
 298
 299             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 300
 301             /* Compute parameters for interactions between i and j atoms */
 302             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 303             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 304
 305             /* REACTION-FIELD ELECTROSTATICS */
 306             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 307             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 308
 309             /* LENNARD-JONES DISPERSION/REPULSION */
 310
 311             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 312             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 313             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 314             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 315             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 316
 317             d                = _fjsp_sub_v2r8(r00,rswitch);
 318             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 319             d2               = _fjsp_mul_v2r8(d,d);
 320             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 321
 322             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 323
 324             /* Evaluate switch function */
 325             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 326             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 327             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 328             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 329
 330             /* Update potential sum for this i atom from the interaction with this j atom. */
 331             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 332             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 333             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 334             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 335             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 336             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 337
 338             fscal            = _fjsp_add_v2r8(felec,fvdw);
 339
 340             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 341
 342             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 343
 344             /* Update vectorial force */
 345             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 346             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 347             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 348
 349             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 350
 351             }
 352
 353             /* Inner loop uses 73 flops */
 354         }
 355
 356         /* End of innermost loop */
 357
 358         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 359                                               f+i_coord_offset,fshift+i_shift_offset);
 360
 361         ggid                        = gid[iidx];
 362         /* Update potential energies */
 363         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 364         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 365
 366         /* Increment number of inner iterations */
 367         inneriter                  += j_index_end - j_index_start;
 368
 369         /* Outer loop uses 9 flops */
 370     }
 371
 372     /* Increment number of outer iterations */
 373     outeriter        += nri;
 374
 375     /* Update outer/inner flops */
 376
 377     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*73);
 378 }
 379 /*
 380  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_F_sparc64_hpc_ace_double
 381  * Electrostatics interaction: ReactionField
 382  * VdW interaction:            LennardJones
 383  * Geometry:                   Particle-Particle
 384  * Calculate force/pot:        Force
 385  */
 386 void
 387 nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_F_sparc64_hpc_ace_double
 388                     (t_nblist                    * gmx_restrict       nlist,
 389                      rvec                        * gmx_restrict          xx,
 390                      rvec                        * gmx_restrict          ff,
 391                      t_forcerec                  * gmx_restrict          fr,
 392                      t_mdatoms                   * gmx_restrict     mdatoms,
 393                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 394                      t_nrnb                      * gmx_restrict        nrnb)
 395 {
 396     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 397      * just 0 for non-waters.
 398      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 399      * jnr indices corresponding to data put in the four positions in the SIMD register.
 400      */
 401     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 402     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 403     int              jnrA,jnrB;
 404     int              j_coord_offsetA,j_coord_offsetB;
 405     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 406     real             rcutoff_scalar;
 407     real             *shiftvec,*fshift,*x,*f;
 408     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 409     int              vdwioffset0;
 410     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 411     int              vdwjidx0A,vdwjidx0B;
 412     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 413     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 414     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 415     real             *charge;
 416     int              nvdwtype;
 417     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 418     int              *vdwtype;
 419     real             *vdwparam;
 420     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 421     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 422     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 423     real             rswitch_scalar,d_scalar;
 424     _fjsp_v2r8       itab_tmp;
 425     _fjsp_v2r8       dummy_mask,cutoff_mask;
 426     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 427     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 428     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 429
 430     x                = xx[0];
 431     f                = ff[0];
 432
 433     nri              = nlist->nri;
 434     iinr             = nlist->iinr;
 435     jindex           = nlist->jindex;
 436     jjnr             = nlist->jjnr;
 437     shiftidx         = nlist->shift;
 438     gid              = nlist->gid;
 439     shiftvec         = fr->shift_vec[0];
 440     fshift           = fr->fshift[0];
 441     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 442     charge           = mdatoms->chargeA;
 443     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 444     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 445     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 446     nvdwtype         = fr->ntype;
 447     vdwparam         = fr->nbfp;
 448     vdwtype          = mdatoms->typeA;
 449
 450     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 451     rcutoff_scalar   = fr->rcoulomb;
 452     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 453     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 454
 455     rswitch_scalar   = fr->rvdw_switch;
 456     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 457     /* Setup switch parameters */
 458     d_scalar         = rcutoff_scalar-rswitch_scalar;
 459     d                = gmx_fjsp_set1_v2r8(d_scalar);
 460     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 461     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 462     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 463     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 464     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 465     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 466
 467     /* Avoid stupid compiler warnings */
 468     jnrA = jnrB = 0;
 469     j_coord_offsetA = 0;
 470     j_coord_offsetB = 0;
 471
 472     outeriter        = 0;
 473     inneriter        = 0;
 474
 475     /* Start outer loop over neighborlists */
 476     for(iidx=0; iidx<nri; iidx++)
 477     {
 478         /* Load shift vector for this list */
 479         i_shift_offset   = DIM*shiftidx[iidx];
 480
 481         /* Load limits for loop over neighbors */
 482         j_index_start    = jindex[iidx];
 483         j_index_end      = jindex[iidx+1];
 484
 485         /* Get outer coordinate index */
 486         inr              = iinr[iidx];
 487         i_coord_offset   = DIM*inr;
 488
 489         /* Load i particle coords and add shift vector */
 490         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 491
 492         fix0             = _fjsp_setzero_v2r8();
 493         fiy0             = _fjsp_setzero_v2r8();
 494         fiz0             = _fjsp_setzero_v2r8();
 495
 496         /* Load parameters for i particles */
 497         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 498         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 499
 500         /* Start inner kernel loop */
 501         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 502         {
 503
 504             /* Get j neighbor index, and coordinate index */
 505             jnrA             = jjnr[jidx];
 506             jnrB             = jjnr[jidx+1];
 507             j_coord_offsetA  = DIM*jnrA;
 508             j_coord_offsetB  = DIM*jnrB;
 509
 510             /* load j atom coordinates */
 511             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 512                                               &jx0,&jy0,&jz0);
 513
 514             /* Calculate displacement vector */
 515             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 516             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 517             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 518
 519             /* Calculate squared distance and things based on it */
 520             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 521
 522             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 523
 524             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 525
 526             /* Load parameters for j particles */
 527             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 528             vdwjidx0A        = 2*vdwtype[jnrA+0];
 529             vdwjidx0B        = 2*vdwtype[jnrB+0];
 530
 531             /**************************
 532              * CALCULATE INTERACTIONS *
 533              **************************/
 534
 535             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 536             {
 537
 538             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 539
 540             /* Compute parameters for interactions between i and j atoms */
 541             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 542             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 543                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 544
 545             /* REACTION-FIELD ELECTROSTATICS */
 546             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 547
 548             /* LENNARD-JONES DISPERSION/REPULSION */
 549
 550             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 551             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 552             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 553             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 554             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 555
 556             d                = _fjsp_sub_v2r8(r00,rswitch);
 557             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 558             d2               = _fjsp_mul_v2r8(d,d);
 559             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 560
 561             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 562
 563             /* Evaluate switch function */
 564             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 565             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 566             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 567
 568             fscal            = _fjsp_add_v2r8(felec,fvdw);
 569
 570             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 571
 572             /* Update vectorial force */
 573             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 574             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 575             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 576
 577             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 578
 579             }
 580
 581             /* Inner loop uses 64 flops */
 582         }
 583
 584         if(jidx<j_index_end)
 585         {
 586
 587             jnrA             = jjnr[jidx];
 588             j_coord_offsetA  = DIM*jnrA;
 589
 590             /* load j atom coordinates */
 591             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 592                                               &jx0,&jy0,&jz0);
 593
 594             /* Calculate displacement vector */
 595             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 596             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 597             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 598
 599             /* Calculate squared distance and things based on it */
 600             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 601
 602             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 603
 604             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 605
 606             /* Load parameters for j particles */
 607             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 608             vdwjidx0A        = 2*vdwtype[jnrA+0];
 609
 610             /**************************
 611              * CALCULATE INTERACTIONS *
 612              **************************/
 613
 614             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 615             {
 616
 617             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 618
 619             /* Compute parameters for interactions between i and j atoms */
 620             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 621             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 622
 623             /* REACTION-FIELD ELECTROSTATICS */
 624             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 625
 626             /* LENNARD-JONES DISPERSION/REPULSION */
 627
 628             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 629             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 630             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 631             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 632             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 633
 634             d                = _fjsp_sub_v2r8(r00,rswitch);
 635             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 636             d2               = _fjsp_mul_v2r8(d,d);
 637             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 638
 639             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 640
 641             /* Evaluate switch function */
 642             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 643             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 644             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 645
 646             fscal            = _fjsp_add_v2r8(felec,fvdw);
 647
 648             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 649
 650             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 651
 652             /* Update vectorial force */
 653             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 654             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 655             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 656
 657             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 658
 659             }
 660
 661             /* Inner loop uses 64 flops */
 662         }
 663
 664         /* End of innermost loop */
 665
 666         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 667                                               f+i_coord_offset,fshift+i_shift_offset);
 668
 669         /* Increment number of inner iterations */
 670         inneriter                  += j_index_end - j_index_start;
 671
 672         /* Outer loop uses 7 flops */
 673     }
 674
 675     /* Increment number of outer iterations */
 676     outeriter        += nri;
 677
 678     /* Update outer/inner flops */
 679
 680     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*64);
 681 }