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 #ifdef HAVE_CONFIG_H
  39 #include <config.h>
  40 #endif
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "types/simple.h"
  46 #include "gromacs/legacyheaders/vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: ReactionField
  54  * VdW interaction:            LennardJones
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_VF_sparc64_hpc_ace_double
  60                     (t_nblist                    * gmx_restrict       nlist,
  61                      rvec                        * gmx_restrict          xx,
  62                      rvec                        * gmx_restrict          ff,
  63                      t_forcerec                  * gmx_restrict          fr,
  64                      t_mdatoms                   * gmx_restrict     mdatoms,
  65                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
  66                      t_nrnb                      * gmx_restrict        nrnb)
  67 {
  68     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  69      * just 0 for non-waters.
  70      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
  71      * jnr indices corresponding to data put in the four positions in the SIMD register.
  72      */
  73     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  74     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  75     int              jnrA,jnrB;
  76     int              j_coord_offsetA,j_coord_offsetB;
  77     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  78     real             rcutoff_scalar;
  79     real             *shiftvec,*fshift,*x,*f;
  80     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  81     int              vdwioffset0;
  82     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  83     int              vdwjidx0A,vdwjidx0B;
  84     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  85     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  86     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
  87     real             *charge;
  88     int              nvdwtype;
  89     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  90     int              *vdwtype;
  91     real             *vdwparam;
  92     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
  93     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
  94     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
  95     real             rswitch_scalar,d_scalar;
  96     _fjsp_v2r8       itab_tmp;
  97     _fjsp_v2r8       dummy_mask,cutoff_mask;
  98     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
  99     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 100     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 101
 102     x                = xx[0];
 103     f                = ff[0];
 104
 105     nri              = nlist->nri;
 106     iinr             = nlist->iinr;
 107     jindex           = nlist->jindex;
 108     jjnr             = nlist->jjnr;
 109     shiftidx         = nlist->shift;
 110     gid              = nlist->gid;
 111     shiftvec         = fr->shift_vec[0];
 112     fshift           = fr->fshift[0];
 113     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 114     charge           = mdatoms->chargeA;
 115     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 116     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 117     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 118     nvdwtype         = fr->ntype;
 119     vdwparam         = fr->nbfp;
 120     vdwtype          = mdatoms->typeA;
 121
 122     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 123     rcutoff_scalar   = fr->rcoulomb;
 124     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 125     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 126
 127     rswitch_scalar   = fr->rvdw_switch;
 128     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 129     /* Setup switch parameters */
 130     d_scalar         = rcutoff_scalar-rswitch_scalar;
 131     d                = gmx_fjsp_set1_v2r8(d_scalar);
 132     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 133     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 134     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 135     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 136     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 137     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 138
 139     /* Avoid stupid compiler warnings */
 140     jnrA = jnrB = 0;
 141     j_coord_offsetA = 0;
 142     j_coord_offsetB = 0;
 143
 144     outeriter        = 0;
 145     inneriter        = 0;
 146
 147     /* Start outer loop over neighborlists */
 148     for(iidx=0; iidx<nri; iidx++)
 149     {
 150         /* Load shift vector for this list */
 151         i_shift_offset   = DIM*shiftidx[iidx];
 152
 153         /* Load limits for loop over neighbors */
 154         j_index_start    = jindex[iidx];
 155         j_index_end      = jindex[iidx+1];
 156
 157         /* Get outer coordinate index */
 158         inr              = iinr[iidx];
 159         i_coord_offset   = DIM*inr;
 160
 161         /* Load i particle coords and add shift vector */
 162         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 163
 164         fix0             = _fjsp_setzero_v2r8();
 165         fiy0             = _fjsp_setzero_v2r8();
 166         fiz0             = _fjsp_setzero_v2r8();
 167
 168         /* Load parameters for i particles */
 169         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 170         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 171
 172         /* Reset potential sums */
 173         velecsum         = _fjsp_setzero_v2r8();
 174         vvdwsum          = _fjsp_setzero_v2r8();
 175
 176         /* Start inner kernel loop */
 177         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 178         {
 179
 180             /* Get j neighbor index, and coordinate index */
 181             jnrA             = jjnr[jidx];
 182             jnrB             = jjnr[jidx+1];
 183             j_coord_offsetA  = DIM*jnrA;
 184             j_coord_offsetB  = DIM*jnrB;
 185
 186             /* load j atom coordinates */
 187             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 188                                               &jx0,&jy0,&jz0);
 189
 190             /* Calculate displacement vector */
 191             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 192             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 193             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 194
 195             /* Calculate squared distance and things based on it */
 196             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 197
 198             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 199
 200             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 201
 202             /* Load parameters for j particles */
 203             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 204             vdwjidx0A        = 2*vdwtype[jnrA+0];
 205             vdwjidx0B        = 2*vdwtype[jnrB+0];
 206
 207             /**************************
 208              * CALCULATE INTERACTIONS *
 209              **************************/
 210
 211             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 212             {
 213
 214             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 215
 216             /* Compute parameters for interactions between i and j atoms */
 217             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 218             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 219                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 220
 221             /* REACTION-FIELD ELECTROSTATICS */
 222             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 223             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 224
 225             /* LENNARD-JONES DISPERSION/REPULSION */
 226
 227             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 228             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 229             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 230             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 231             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 232
 233             d                = _fjsp_sub_v2r8(r00,rswitch);
 234             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 235             d2               = _fjsp_mul_v2r8(d,d);
 236             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 237
 238             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 239
 240             /* Evaluate switch function */
 241             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 242             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 243             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 244             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 245
 246             /* Update potential sum for this i atom from the interaction with this j atom. */
 247             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 248             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 249             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 250             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 251
 252             fscal            = _fjsp_add_v2r8(felec,fvdw);
 253
 254             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 255
 256             /* Update vectorial force */
 257             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 258             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 259             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 260
 261             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 262
 263             }
 264
 265             /* Inner loop uses 73 flops */
 266         }
 267
 268         if(jidx<j_index_end)
 269         {
 270
 271             jnrA             = jjnr[jidx];
 272             j_coord_offsetA  = DIM*jnrA;
 273
 274             /* load j atom coordinates */
 275             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 276                                               &jx0,&jy0,&jz0);
 277
 278             /* Calculate displacement vector */
 279             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 280             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 281             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 282
 283             /* Calculate squared distance and things based on it */
 284             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 285
 286             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 287
 288             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 289
 290             /* Load parameters for j particles */
 291             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 292             vdwjidx0A        = 2*vdwtype[jnrA+0];
 293
 294             /**************************
 295              * CALCULATE INTERACTIONS *
 296              **************************/
 297
 298             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 299             {
 300
 301             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 302
 303             /* Compute parameters for interactions between i and j atoms */
 304             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 305             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 306                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 307
 308             /* REACTION-FIELD ELECTROSTATICS */
 309             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(_fjsp_madd_v2r8(krf,rsq00,rinv00),crf));
 310             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 311
 312             /* LENNARD-JONES DISPERSION/REPULSION */
 313
 314             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 315             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 316             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 317             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 318             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 319
 320             d                = _fjsp_sub_v2r8(r00,rswitch);
 321             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 322             d2               = _fjsp_mul_v2r8(d,d);
 323             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 324
 325             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 326
 327             /* Evaluate switch function */
 328             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 329             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 330             vvdw             = _fjsp_mul_v2r8(vvdw,sw);
 331             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 332
 333             /* Update potential sum for this i atom from the interaction with this j atom. */
 334             velec            = _fjsp_and_v2r8(velec,cutoff_mask);
 335             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 336             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 337             vvdw             = _fjsp_and_v2r8(vvdw,cutoff_mask);
 338             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 339             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 340
 341             fscal            = _fjsp_add_v2r8(felec,fvdw);
 342
 343             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 344
 345             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 346
 347             /* Update vectorial force */
 348             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 349             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 350             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 351
 352             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 353
 354             }
 355
 356             /* Inner loop uses 73 flops */
 357         }
 358
 359         /* End of innermost loop */
 360
 361         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 362                                               f+i_coord_offset,fshift+i_shift_offset);
 363
 364         ggid                        = gid[iidx];
 365         /* Update potential energies */
 366         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 367         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 368
 369         /* Increment number of inner iterations */
 370         inneriter                  += j_index_end - j_index_start;
 371
 372         /* Outer loop uses 9 flops */
 373     }
 374
 375     /* Increment number of outer iterations */
 376     outeriter        += nri;
 377
 378     /* Update outer/inner flops */
 379
 380     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*73);
 381 }
 382 /*
 383  * Gromacs nonbonded kernel:   nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_F_sparc64_hpc_ace_double
 384  * Electrostatics interaction: ReactionField
 385  * VdW interaction:            LennardJones
 386  * Geometry:                   Particle-Particle
 387  * Calculate force/pot:        Force
 388  */
 389 void
 390 nb_kernel_ElecRFCut_VdwLJSw_GeomP1P1_F_sparc64_hpc_ace_double
 391                     (t_nblist                    * gmx_restrict       nlist,
 392                      rvec                        * gmx_restrict          xx,
 393                      rvec                        * gmx_restrict          ff,
 394                      t_forcerec                  * gmx_restrict          fr,
 395                      t_mdatoms                   * gmx_restrict     mdatoms,
 396                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 397                      t_nrnb                      * gmx_restrict        nrnb)
 398 {
 399     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 400      * just 0 for non-waters.
 401      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 402      * jnr indices corresponding to data put in the four positions in the SIMD register.
 403      */
 404     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 405     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 406     int              jnrA,jnrB;
 407     int              j_coord_offsetA,j_coord_offsetB;
 408     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 409     real             rcutoff_scalar;
 410     real             *shiftvec,*fshift,*x,*f;
 411     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 412     int              vdwioffset0;
 413     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 414     int              vdwjidx0A,vdwjidx0B;
 415     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 416     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 417     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 418     real             *charge;
 419     int              nvdwtype;
 420     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 421     int              *vdwtype;
 422     real             *vdwparam;
 423     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 424     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 425     _fjsp_v2r8       rswitch,swV3,swV4,swV5,swF2,swF3,swF4,d,d2,sw,dsw;
 426     real             rswitch_scalar,d_scalar;
 427     _fjsp_v2r8       itab_tmp;
 428     _fjsp_v2r8       dummy_mask,cutoff_mask;
 429     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 430     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 431     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 432
 433     x                = xx[0];
 434     f                = ff[0];
 435
 436     nri              = nlist->nri;
 437     iinr             = nlist->iinr;
 438     jindex           = nlist->jindex;
 439     jjnr             = nlist->jjnr;
 440     shiftidx         = nlist->shift;
 441     gid              = nlist->gid;
 442     shiftvec         = fr->shift_vec[0];
 443     fshift           = fr->fshift[0];
 444     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 445     charge           = mdatoms->chargeA;
 446     krf              = gmx_fjsp_set1_v2r8(fr->ic->k_rf);
 447     krf2             = gmx_fjsp_set1_v2r8(fr->ic->k_rf*2.0);
 448     crf              = gmx_fjsp_set1_v2r8(fr->ic->c_rf);
 449     nvdwtype         = fr->ntype;
 450     vdwparam         = fr->nbfp;
 451     vdwtype          = mdatoms->typeA;
 452
 453     /* When we use explicit cutoffs the value must be identical for elec and VdW, so use elec as an arbitrary choice */
 454     rcutoff_scalar   = fr->rcoulomb;
 455     rcutoff          = gmx_fjsp_set1_v2r8(rcutoff_scalar);
 456     rcutoff2         = _fjsp_mul_v2r8(rcutoff,rcutoff);
 457
 458     rswitch_scalar   = fr->rvdw_switch;
 459     rswitch          = gmx_fjsp_set1_v2r8(rswitch_scalar);
 460     /* Setup switch parameters */
 461     d_scalar         = rcutoff_scalar-rswitch_scalar;
 462     d                = gmx_fjsp_set1_v2r8(d_scalar);
 463     swV3             = gmx_fjsp_set1_v2r8(-10.0/(d_scalar*d_scalar*d_scalar));
 464     swV4             = gmx_fjsp_set1_v2r8( 15.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 465     swV5             = gmx_fjsp_set1_v2r8( -6.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 466     swF2             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar));
 467     swF3             = gmx_fjsp_set1_v2r8( 60.0/(d_scalar*d_scalar*d_scalar*d_scalar));
 468     swF4             = gmx_fjsp_set1_v2r8(-30.0/(d_scalar*d_scalar*d_scalar*d_scalar*d_scalar));
 469
 470     /* Avoid stupid compiler warnings */
 471     jnrA = jnrB = 0;
 472     j_coord_offsetA = 0;
 473     j_coord_offsetB = 0;
 474
 475     outeriter        = 0;
 476     inneriter        = 0;
 477
 478     /* Start outer loop over neighborlists */
 479     for(iidx=0; iidx<nri; iidx++)
 480     {
 481         /* Load shift vector for this list */
 482         i_shift_offset   = DIM*shiftidx[iidx];
 483
 484         /* Load limits for loop over neighbors */
 485         j_index_start    = jindex[iidx];
 486         j_index_end      = jindex[iidx+1];
 487
 488         /* Get outer coordinate index */
 489         inr              = iinr[iidx];
 490         i_coord_offset   = DIM*inr;
 491
 492         /* Load i particle coords and add shift vector */
 493         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 494
 495         fix0             = _fjsp_setzero_v2r8();
 496         fiy0             = _fjsp_setzero_v2r8();
 497         fiz0             = _fjsp_setzero_v2r8();
 498
 499         /* Load parameters for i particles */
 500         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 501         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 502
 503         /* Start inner kernel loop */
 504         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 505         {
 506
 507             /* Get j neighbor index, and coordinate index */
 508             jnrA             = jjnr[jidx];
 509             jnrB             = jjnr[jidx+1];
 510             j_coord_offsetA  = DIM*jnrA;
 511             j_coord_offsetB  = DIM*jnrB;
 512
 513             /* load j atom coordinates */
 514             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 515                                               &jx0,&jy0,&jz0);
 516
 517             /* Calculate displacement vector */
 518             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 519             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 520             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 521
 522             /* Calculate squared distance and things based on it */
 523             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 524
 525             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 526
 527             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 528
 529             /* Load parameters for j particles */
 530             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 531             vdwjidx0A        = 2*vdwtype[jnrA+0];
 532             vdwjidx0B        = 2*vdwtype[jnrB+0];
 533
 534             /**************************
 535              * CALCULATE INTERACTIONS *
 536              **************************/
 537
 538             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 539             {
 540
 541             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 542
 543             /* Compute parameters for interactions between i and j atoms */
 544             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 545             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 546                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 547
 548             /* REACTION-FIELD ELECTROSTATICS */
 549             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 550
 551             /* LENNARD-JONES DISPERSION/REPULSION */
 552
 553             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 554             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 555             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 556             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 557             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 558
 559             d                = _fjsp_sub_v2r8(r00,rswitch);
 560             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 561             d2               = _fjsp_mul_v2r8(d,d);
 562             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 563
 564             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 565
 566             /* Evaluate switch function */
 567             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 568             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 569             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 570
 571             fscal            = _fjsp_add_v2r8(felec,fvdw);
 572
 573             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 574
 575             /* Update vectorial force */
 576             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 577             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 578             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 579
 580             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 581
 582             }
 583
 584             /* Inner loop uses 64 flops */
 585         }
 586
 587         if(jidx<j_index_end)
 588         {
 589
 590             jnrA             = jjnr[jidx];
 591             j_coord_offsetA  = DIM*jnrA;
 592
 593             /* load j atom coordinates */
 594             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 595                                               &jx0,&jy0,&jz0);
 596
 597             /* Calculate displacement vector */
 598             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 599             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 600             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 601
 602             /* Calculate squared distance and things based on it */
 603             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 604
 605             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 606
 607             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 608
 609             /* Load parameters for j particles */
 610             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 611             vdwjidx0A        = 2*vdwtype[jnrA+0];
 612
 613             /**************************
 614              * CALCULATE INTERACTIONS *
 615              **************************/
 616
 617             if (gmx_fjsp_any_lt_v2r8(rsq00,rcutoff2))
 618             {
 619
 620             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 621
 622             /* Compute parameters for interactions between i and j atoms */
 623             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 624             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 625                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 626
 627             /* REACTION-FIELD ELECTROSTATICS */
 628             felec            = _fjsp_mul_v2r8(qq00,_fjsp_msub_v2r8(rinv00,rinvsq00,krf2));
 629
 630             /* LENNARD-JONES DISPERSION/REPULSION */
 631
 632             rinvsix          = _fjsp_mul_v2r8(_fjsp_mul_v2r8(rinvsq00,rinvsq00),rinvsq00);
 633             vvdw6            = _fjsp_mul_v2r8(c6_00,rinvsix);
 634             vvdw12           = _fjsp_mul_v2r8(c12_00,_fjsp_mul_v2r8(rinvsix,rinvsix));
 635             vvdw             = _fjsp_msub_v2r8( vvdw12,one_twelfth, _fjsp_mul_v2r8(vvdw6,one_sixth) );
 636             fvdw             = _fjsp_mul_v2r8(_fjsp_sub_v2r8(vvdw12,vvdw6),rinvsq00);
 637
 638             d                = _fjsp_sub_v2r8(r00,rswitch);
 639             d                = _fjsp_max_v2r8(d,_fjsp_setzero_v2r8());
 640             d2               = _fjsp_mul_v2r8(d,d);
 641             sw               = _fjsp_add_v2r8(one,_fjsp_mul_v2r8(d2,_fjsp_mul_v2r8(d,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swV5,swV4),swV3))));
 642
 643             dsw              = _fjsp_mul_v2r8(d2,_fjsp_madd_v2r8(d,_fjsp_madd_v2r8(d,swF4,swF3),swF2));
 644
 645             /* Evaluate switch function */
 646             /* fscal'=f'/r=-(v*sw)'/r=-(v'*sw+v*dsw)/r=-v'*sw/r-v*dsw/r=fscal*sw-v*dsw/r */
 647             fvdw             = _fjsp_msub_v2r8( fvdw,sw , _fjsp_mul_v2r8(rinv00,_fjsp_mul_v2r8(vvdw,dsw)) );
 648             cutoff_mask      = _fjsp_cmplt_v2r8(rsq00,rcutoff2);
 649
 650             fscal            = _fjsp_add_v2r8(felec,fvdw);
 651
 652             fscal            = _fjsp_and_v2r8(fscal,cutoff_mask);
 653
 654             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 655
 656             /* Update vectorial force */
 657             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 658             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 659             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 660
 661             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 662
 663             }
 664
 665             /* Inner loop uses 64 flops */
 666         }
 667
 668         /* End of innermost loop */
 669
 670         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 671                                               f+i_coord_offset,fshift+i_shift_offset);
 672
 673         /* Increment number of inner iterations */
 674         inneriter                  += j_index_end - j_index_start;
 675
 676         /* Outer loop uses 7 flops */
 677     }
 678
 679     /* Increment number of outer iterations */
 680     outeriter        += nri;
 681
 682     /* Update outer/inner flops */
 683
 684     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*64);
 685 }