src/gromacs/gmxlib/nonbonded/nb_kernel_sparc64_hpc_ace_double/nb_kernel_ElecEw_VdwCSTab_GeomP1P1_sparc64_hpc_ace_double.c

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013, 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 "vec.h"
  47 #include "nrnb.h"
  48
  49 #include "kernelutil_sparc64_hpc_ace_double.h"
  50
  51 /*
  52  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomP1P1_VF_sparc64_hpc_ace_double
  53  * Electrostatics interaction: Ewald
  54  * VdW interaction:            CubicSplineTable
  55  * Geometry:                   Particle-Particle
  56  * Calculate force/pot:        PotentialAndForce
  57  */
  58 void
  59 nb_kernel_ElecEw_VdwCSTab_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       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
  95     real             *vftab;
  96     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
  97     real             *ewtab;
  98     _fjsp_v2r8       itab_tmp;
  99     _fjsp_v2r8       dummy_mask,cutoff_mask;
 100     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 101     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 102     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 103
 104     x                = xx[0];
 105     f                = ff[0];
 106
 107     nri              = nlist->nri;
 108     iinr             = nlist->iinr;
 109     jindex           = nlist->jindex;
 110     jjnr             = nlist->jjnr;
 111     shiftidx         = nlist->shift;
 112     gid              = nlist->gid;
 113     shiftvec         = fr->shift_vec[0];
 114     fshift           = fr->fshift[0];
 115     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 116     charge           = mdatoms->chargeA;
 117     nvdwtype         = fr->ntype;
 118     vdwparam         = fr->nbfp;
 119     vdwtype          = mdatoms->typeA;
 120
 121     vftab            = kernel_data->table_vdw->data;
 122     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 123
 124     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 125     ewtab            = fr->ic->tabq_coul_FDV0;
 126     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 127     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 128
 129     /* Avoid stupid compiler warnings */
 130     jnrA = jnrB = 0;
 131     j_coord_offsetA = 0;
 132     j_coord_offsetB = 0;
 133
 134     outeriter        = 0;
 135     inneriter        = 0;
 136
 137     /* Start outer loop over neighborlists */
 138     for(iidx=0; iidx<nri; iidx++)
 139     {
 140         /* Load shift vector for this list */
 141         i_shift_offset   = DIM*shiftidx[iidx];
 142
 143         /* Load limits for loop over neighbors */
 144         j_index_start    = jindex[iidx];
 145         j_index_end      = jindex[iidx+1];
 146
 147         /* Get outer coordinate index */
 148         inr              = iinr[iidx];
 149         i_coord_offset   = DIM*inr;
 150
 151         /* Load i particle coords and add shift vector */
 152         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 153
 154         fix0             = _fjsp_setzero_v2r8();
 155         fiy0             = _fjsp_setzero_v2r8();
 156         fiz0             = _fjsp_setzero_v2r8();
 157
 158         /* Load parameters for i particles */
 159         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 160         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 161
 162         /* Reset potential sums */
 163         velecsum         = _fjsp_setzero_v2r8();
 164         vvdwsum          = _fjsp_setzero_v2r8();
 165
 166         /* Start inner kernel loop */
 167         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 168         {
 169
 170             /* Get j neighbor index, and coordinate index */
 171             jnrA             = jjnr[jidx];
 172             jnrB             = jjnr[jidx+1];
 173             j_coord_offsetA  = DIM*jnrA;
 174             j_coord_offsetB  = DIM*jnrB;
 175
 176             /* load j atom coordinates */
 177             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 178                                               &jx0,&jy0,&jz0);
 179
 180             /* Calculate displacement vector */
 181             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 182             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 183             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 184
 185             /* Calculate squared distance and things based on it */
 186             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 187
 188             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 189
 190             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 191
 192             /* Load parameters for j particles */
 193             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 194             vdwjidx0A        = 2*vdwtype[jnrA+0];
 195             vdwjidx0B        = 2*vdwtype[jnrB+0];
 196
 197             /**************************
 198              * CALCULATE INTERACTIONS *
 199              **************************/
 200
 201             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 202
 203             /* Compute parameters for interactions between i and j atoms */
 204             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 205             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 206                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 207
 208             /* Calculate table index by multiplying r with table scale and truncate to integer */
 209             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 210             itab_tmp         = _fjsp_dtox_v2r8(rt);
 211             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 212             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 213             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 214
 215             vfconv.i[0]     *= 8;
 216             vfconv.i[1]     *= 8;
 217
 218             /* EWALD ELECTROSTATICS */
 219
 220             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 221             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 222             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 223             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 224             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 225
 226             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 227             ewtabD           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[1] );
 228             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 229             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 230             ewtabFn          = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[1] +2);
 231             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 232             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 233             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 234             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 235             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 236
 237             /* CUBIC SPLINE TABLE DISPERSION */
 238             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 239             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 240             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 241             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 242             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 243             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 244             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 245             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 246             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 247             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 248             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 249
 250             /* CUBIC SPLINE TABLE REPULSION */
 251             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 252             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 253             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 254             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 255             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 256             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 257             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 258             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 259             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 260             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 261             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 262             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 263             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 264
 265             /* Update potential sum for this i atom from the interaction with this j atom. */
 266             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 267             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 268
 269             fscal            = _fjsp_add_v2r8(felec,fvdw);
 270
 271             /* Update vectorial force */
 272             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 273             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 274             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 275
 276             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 277
 278             /* Inner loop uses 78 flops */
 279         }
 280
 281         if(jidx<j_index_end)
 282         {
 283
 284             jnrA             = jjnr[jidx];
 285             j_coord_offsetA  = DIM*jnrA;
 286
 287             /* load j atom coordinates */
 288             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 289                                               &jx0,&jy0,&jz0);
 290
 291             /* Calculate displacement vector */
 292             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 293             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 294             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 295
 296             /* Calculate squared distance and things based on it */
 297             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 298
 299             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 300
 301             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 302
 303             /* Load parameters for j particles */
 304             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 305             vdwjidx0A        = 2*vdwtype[jnrA+0];
 306
 307             /**************************
 308              * CALCULATE INTERACTIONS *
 309              **************************/
 310
 311             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 312
 313             /* Compute parameters for interactions between i and j atoms */
 314             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 315             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 316
 317             /* Calculate table index by multiplying r with table scale and truncate to integer */
 318             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 319             itab_tmp         = _fjsp_dtox_v2r8(rt);
 320             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 321             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 322             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 323
 324             vfconv.i[0]     *= 8;
 325             vfconv.i[1]     *= 8;
 326
 327             /* EWALD ELECTROSTATICS */
 328
 329             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 330             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 331             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 332             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 333             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 334
 335             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 336             ewtabD           = _fjsp_setzero_v2r8();
 337             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 338             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 339             ewtabFn          = _fjsp_setzero_v2r8();
 340             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 341             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 342             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 343             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 344             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 345
 346             /* CUBIC SPLINE TABLE DISPERSION */
 347             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 348             F                = _fjsp_setzero_v2r8();
 349             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 350             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 351             H                = _fjsp_setzero_v2r8();
 352             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 353             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 354             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 355             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 356             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 357             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 358
 359             /* CUBIC SPLINE TABLE REPULSION */
 360             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 361             F                = _fjsp_setzero_v2r8();
 362             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 363             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 364             H                = _fjsp_setzero_v2r8();
 365             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 366             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 367             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 368             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 369             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 370             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 371             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 372             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 373
 374             /* Update potential sum for this i atom from the interaction with this j atom. */
 375             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 376             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 377             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 378             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 379
 380             fscal            = _fjsp_add_v2r8(felec,fvdw);
 381
 382             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 383
 384             /* Update vectorial force */
 385             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 386             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 387             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 388
 389             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 390
 391             /* Inner loop uses 78 flops */
 392         }
 393
 394         /* End of innermost loop */
 395
 396         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 397                                               f+i_coord_offset,fshift+i_shift_offset);
 398
 399         ggid                        = gid[iidx];
 400         /* Update potential energies */
 401         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 402         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 403
 404         /* Increment number of inner iterations */
 405         inneriter                  += j_index_end - j_index_start;
 406
 407         /* Outer loop uses 9 flops */
 408     }
 409
 410     /* Increment number of outer iterations */
 411     outeriter        += nri;
 412
 413     /* Update outer/inner flops */
 414
 415     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*78);
 416 }
 417 /*
 418  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 419  * Electrostatics interaction: Ewald
 420  * VdW interaction:            CubicSplineTable
 421  * Geometry:                   Particle-Particle
 422  * Calculate force/pot:        Force
 423  */
 424 void
 425 nb_kernel_ElecEw_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 426                     (t_nblist                    * gmx_restrict       nlist,
 427                      rvec                        * gmx_restrict          xx,
 428                      rvec                        * gmx_restrict          ff,
 429                      t_forcerec                  * gmx_restrict          fr,
 430                      t_mdatoms                   * gmx_restrict     mdatoms,
 431                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 432                      t_nrnb                      * gmx_restrict        nrnb)
 433 {
 434     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 435      * just 0 for non-waters.
 436      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 437      * jnr indices corresponding to data put in the four positions in the SIMD register.
 438      */
 439     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 440     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 441     int              jnrA,jnrB;
 442     int              j_coord_offsetA,j_coord_offsetB;
 443     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 444     real             rcutoff_scalar;
 445     real             *shiftvec,*fshift,*x,*f;
 446     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 447     int              vdwioffset0;
 448     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 449     int              vdwjidx0A,vdwjidx0B;
 450     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 451     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 452     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 453     real             *charge;
 454     int              nvdwtype;
 455     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 456     int              *vdwtype;
 457     real             *vdwparam;
 458     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 459     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 460     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 461     real             *vftab;
 462     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 463     real             *ewtab;
 464     _fjsp_v2r8       itab_tmp;
 465     _fjsp_v2r8       dummy_mask,cutoff_mask;
 466     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 467     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 468     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 469
 470     x                = xx[0];
 471     f                = ff[0];
 472
 473     nri              = nlist->nri;
 474     iinr             = nlist->iinr;
 475     jindex           = nlist->jindex;
 476     jjnr             = nlist->jjnr;
 477     shiftidx         = nlist->shift;
 478     gid              = nlist->gid;
 479     shiftvec         = fr->shift_vec[0];
 480     fshift           = fr->fshift[0];
 481     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 482     charge           = mdatoms->chargeA;
 483     nvdwtype         = fr->ntype;
 484     vdwparam         = fr->nbfp;
 485     vdwtype          = mdatoms->typeA;
 486
 487     vftab            = kernel_data->table_vdw->data;
 488     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 489
 490     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 491     ewtab            = fr->ic->tabq_coul_F;
 492     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 493     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 494
 495     /* Avoid stupid compiler warnings */
 496     jnrA = jnrB = 0;
 497     j_coord_offsetA = 0;
 498     j_coord_offsetB = 0;
 499
 500     outeriter        = 0;
 501     inneriter        = 0;
 502
 503     /* Start outer loop over neighborlists */
 504     for(iidx=0; iidx<nri; iidx++)
 505     {
 506         /* Load shift vector for this list */
 507         i_shift_offset   = DIM*shiftidx[iidx];
 508
 509         /* Load limits for loop over neighbors */
 510         j_index_start    = jindex[iidx];
 511         j_index_end      = jindex[iidx+1];
 512
 513         /* Get outer coordinate index */
 514         inr              = iinr[iidx];
 515         i_coord_offset   = DIM*inr;
 516
 517         /* Load i particle coords and add shift vector */
 518         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 519
 520         fix0             = _fjsp_setzero_v2r8();
 521         fiy0             = _fjsp_setzero_v2r8();
 522         fiz0             = _fjsp_setzero_v2r8();
 523
 524         /* Load parameters for i particles */
 525         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 526         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 527
 528         /* Start inner kernel loop */
 529         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 530         {
 531
 532             /* Get j neighbor index, and coordinate index */
 533             jnrA             = jjnr[jidx];
 534             jnrB             = jjnr[jidx+1];
 535             j_coord_offsetA  = DIM*jnrA;
 536             j_coord_offsetB  = DIM*jnrB;
 537
 538             /* load j atom coordinates */
 539             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 540                                               &jx0,&jy0,&jz0);
 541
 542             /* Calculate displacement vector */
 543             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 544             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 545             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 546
 547             /* Calculate squared distance and things based on it */
 548             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 549
 550             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 551
 552             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 553
 554             /* Load parameters for j particles */
 555             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 556             vdwjidx0A        = 2*vdwtype[jnrA+0];
 557             vdwjidx0B        = 2*vdwtype[jnrB+0];
 558
 559             /**************************
 560              * CALCULATE INTERACTIONS *
 561              **************************/
 562
 563             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 564
 565             /* Compute parameters for interactions between i and j atoms */
 566             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 567             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 568                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 569
 570             /* Calculate table index by multiplying r with table scale and truncate to integer */
 571             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 572             itab_tmp         = _fjsp_dtox_v2r8(rt);
 573             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 574             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 575             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 576
 577             vfconv.i[0]     *= 8;
 578             vfconv.i[1]     *= 8;
 579
 580             /* EWALD ELECTROSTATICS */
 581
 582             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 583             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 584             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 585             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 586             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 587
 588             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 589                                          &ewtabF,&ewtabFn);
 590             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 591             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 592
 593             /* CUBIC SPLINE TABLE DISPERSION */
 594             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 595             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 596             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 597             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 598             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 599             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 600             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 601             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 602             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 603
 604             /* CUBIC SPLINE TABLE REPULSION */
 605             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 606             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 607             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 608             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 609             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 610             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 611             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 612             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 613             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 614             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 615
 616             fscal            = _fjsp_add_v2r8(felec,fvdw);
 617
 618             /* Update vectorial force */
 619             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 620             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 621             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 622
 623             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 624
 625             /* Inner loop uses 65 flops */
 626         }
 627
 628         if(jidx<j_index_end)
 629         {
 630
 631             jnrA             = jjnr[jidx];
 632             j_coord_offsetA  = DIM*jnrA;
 633
 634             /* load j atom coordinates */
 635             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 636                                               &jx0,&jy0,&jz0);
 637
 638             /* Calculate displacement vector */
 639             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 640             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 641             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 642
 643             /* Calculate squared distance and things based on it */
 644             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 645
 646             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 647
 648             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 649
 650             /* Load parameters for j particles */
 651             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 652             vdwjidx0A        = 2*vdwtype[jnrA+0];
 653
 654             /**************************
 655              * CALCULATE INTERACTIONS *
 656              **************************/
 657
 658             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 659
 660             /* Compute parameters for interactions between i and j atoms */
 661             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 662             gmx_fjsp_load_1pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 663
 664             /* Calculate table index by multiplying r with table scale and truncate to integer */
 665             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 666             itab_tmp         = _fjsp_dtox_v2r8(rt);
 667             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 668             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 669             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 670
 671             vfconv.i[0]     *= 8;
 672             vfconv.i[1]     *= 8;
 673
 674             /* EWALD ELECTROSTATICS */
 675
 676             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 677             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 678             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 679             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 680             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 681
 682             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 683             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 684             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 685
 686             /* CUBIC SPLINE TABLE DISPERSION */
 687             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 688             F                = _fjsp_setzero_v2r8();
 689             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 690             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 691             H                = _fjsp_setzero_v2r8();
 692             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 693             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 694             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 695             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 696
 697             /* CUBIC SPLINE TABLE REPULSION */
 698             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 699             F                = _fjsp_setzero_v2r8();
 700             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 701             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 702             H                = _fjsp_setzero_v2r8();
 703             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 704             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 705             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 706             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 707             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 708
 709             fscal            = _fjsp_add_v2r8(felec,fvdw);
 710
 711             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 712
 713             /* Update vectorial force */
 714             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 715             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 716             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 717
 718             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 719
 720             /* Inner loop uses 65 flops */
 721         }
 722
 723         /* End of innermost loop */
 724
 725         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 726                                               f+i_coord_offset,fshift+i_shift_offset);
 727
 728         /* Increment number of inner iterations */
 729         inneriter                  += j_index_end - j_index_start;
 730
 731         /* Outer loop uses 7 flops */
 732     }
 733
 734     /* Increment number of outer iterations */
 735     outeriter        += nri;
 736
 737     /* Update outer/inner flops */
 738
 739     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*65);
 740 }