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) 2012,2013,2014, 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 "gmxpre.h"
  39
  40 #include "config.h"
  41
  42 #include <math.h>
  43
  44 #include "../nb_kernel.h"
  45 #include "gromacs/legacyheaders/types/simple.h"
  46 #include "gromacs/math/vec.h"
  47 #include "gromacs/legacyheaders/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_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 316                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 317
 318             /* Calculate table index by multiplying r with table scale and truncate to integer */
 319             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 320             itab_tmp         = _fjsp_dtox_v2r8(rt);
 321             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 322             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 323             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 324
 325             vfconv.i[0]     *= 8;
 326             vfconv.i[1]     *= 8;
 327
 328             /* EWALD ELECTROSTATICS */
 329
 330             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 331             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 332             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 333             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 334             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 335
 336             ewtabF           = _fjsp_load_v2r8( ewtab + 4*ewconv.i[0] );
 337             ewtabD           = _fjsp_setzero_v2r8();
 338             GMX_FJSP_TRANSPOSE2_V2R8(ewtabF,ewtabD);
 339             ewtabV           = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(), ewtab + 4*ewconv.i[0] +2);
 340             ewtabFn          = _fjsp_setzero_v2r8();
 341             GMX_FJSP_TRANSPOSE2_V2R8(ewtabV,ewtabFn);
 342             felec            = _fjsp_madd_v2r8(eweps,ewtabD,ewtabF);
 343             velec            = _fjsp_nmsub_v2r8(_fjsp_mul_v2r8(ewtabhalfspace,eweps) ,_fjsp_add_v2r8(ewtabF,felec), ewtabV);
 344             velec            = _fjsp_mul_v2r8(qq00,_fjsp_sub_v2r8(rinv00,velec));
 345             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 346
 347             /* CUBIC SPLINE TABLE DISPERSION */
 348             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 349             F                = _fjsp_setzero_v2r8();
 350             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 351             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 352             H                = _fjsp_setzero_v2r8();
 353             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 354             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 355             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 356             vvdw6            = _fjsp_mul_v2r8(c6_00,VV);
 357             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 358             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 359
 360             /* CUBIC SPLINE TABLE REPULSION */
 361             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 362             F                = _fjsp_setzero_v2r8();
 363             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 364             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 365             H                = _fjsp_setzero_v2r8();
 366             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 367             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 368             VV               = _fjsp_madd_v2r8(vfeps,Fp,Y);
 369             vvdw12           = _fjsp_mul_v2r8(c12_00,VV);
 370             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 371             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 372             vvdw             = _fjsp_add_v2r8(vvdw12,vvdw6);
 373             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 374
 375             /* Update potential sum for this i atom from the interaction with this j atom. */
 376             velec            = _fjsp_unpacklo_v2r8(velec,_fjsp_setzero_v2r8());
 377             velecsum         = _fjsp_add_v2r8(velecsum,velec);
 378             vvdw             = _fjsp_unpacklo_v2r8(vvdw,_fjsp_setzero_v2r8());
 379             vvdwsum          = _fjsp_add_v2r8(vvdwsum,vvdw);
 380
 381             fscal            = _fjsp_add_v2r8(felec,fvdw);
 382
 383             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 384
 385             /* Update vectorial force */
 386             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 387             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 388             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 389
 390             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 391
 392             /* Inner loop uses 78 flops */
 393         }
 394
 395         /* End of innermost loop */
 396
 397         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 398                                               f+i_coord_offset,fshift+i_shift_offset);
 399
 400         ggid                        = gid[iidx];
 401         /* Update potential energies */
 402         gmx_fjsp_update_1pot_v2r8(velecsum,kernel_data->energygrp_elec+ggid);
 403         gmx_fjsp_update_1pot_v2r8(vvdwsum,kernel_data->energygrp_vdw+ggid);
 404
 405         /* Increment number of inner iterations */
 406         inneriter                  += j_index_end - j_index_start;
 407
 408         /* Outer loop uses 9 flops */
 409     }
 410
 411     /* Increment number of outer iterations */
 412     outeriter        += nri;
 413
 414     /* Update outer/inner flops */
 415
 416     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*78);
 417 }
 418 /*
 419  * Gromacs nonbonded kernel:   nb_kernel_ElecEw_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 420  * Electrostatics interaction: Ewald
 421  * VdW interaction:            CubicSplineTable
 422  * Geometry:                   Particle-Particle
 423  * Calculate force/pot:        Force
 424  */
 425 void
 426 nb_kernel_ElecEw_VdwCSTab_GeomP1P1_F_sparc64_hpc_ace_double
 427                     (t_nblist                    * gmx_restrict       nlist,
 428                      rvec                        * gmx_restrict          xx,
 429                      rvec                        * gmx_restrict          ff,
 430                      t_forcerec                  * gmx_restrict          fr,
 431                      t_mdatoms                   * gmx_restrict     mdatoms,
 432                      nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
 433                      t_nrnb                      * gmx_restrict        nrnb)
 434 {
 435     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 436      * just 0 for non-waters.
 437      * Suffixes A,B refer to j loop unrolling done with double precision SIMD, e.g. for the two different
 438      * jnr indices corresponding to data put in the four positions in the SIMD register.
 439      */
 440     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 441     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 442     int              jnrA,jnrB;
 443     int              j_coord_offsetA,j_coord_offsetB;
 444     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 445     real             rcutoff_scalar;
 446     real             *shiftvec,*fshift,*x,*f;
 447     _fjsp_v2r8       tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 448     int              vdwioffset0;
 449     _fjsp_v2r8       ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 450     int              vdwjidx0A,vdwjidx0B;
 451     _fjsp_v2r8       jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 452     _fjsp_v2r8       dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 453     _fjsp_v2r8       velec,felec,velecsum,facel,crf,krf,krf2;
 454     real             *charge;
 455     int              nvdwtype;
 456     _fjsp_v2r8       rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 457     int              *vdwtype;
 458     real             *vdwparam;
 459     _fjsp_v2r8       one_sixth   = gmx_fjsp_set1_v2r8(1.0/6.0);
 460     _fjsp_v2r8       one_twelfth = gmx_fjsp_set1_v2r8(1.0/12.0);
 461     _fjsp_v2r8       rt,vfeps,vftabscale,Y,F,G,H,Heps,Fp,VV,FF,twovfeps;
 462     real             *vftab;
 463     _fjsp_v2r8       ewtabscale,eweps,sh_ewald,ewrt,ewtabhalfspace,ewtabF,ewtabFn,ewtabD,ewtabV;
 464     real             *ewtab;
 465     _fjsp_v2r8       itab_tmp;
 466     _fjsp_v2r8       dummy_mask,cutoff_mask;
 467     _fjsp_v2r8       one     = gmx_fjsp_set1_v2r8(1.0);
 468     _fjsp_v2r8       two     = gmx_fjsp_set1_v2r8(2.0);
 469     union { _fjsp_v2r8 simd; long long int i[2]; } vfconv,gbconv,ewconv;
 470
 471     x                = xx[0];
 472     f                = ff[0];
 473
 474     nri              = nlist->nri;
 475     iinr             = nlist->iinr;
 476     jindex           = nlist->jindex;
 477     jjnr             = nlist->jjnr;
 478     shiftidx         = nlist->shift;
 479     gid              = nlist->gid;
 480     shiftvec         = fr->shift_vec[0];
 481     fshift           = fr->fshift[0];
 482     facel            = gmx_fjsp_set1_v2r8(fr->epsfac);
 483     charge           = mdatoms->chargeA;
 484     nvdwtype         = fr->ntype;
 485     vdwparam         = fr->nbfp;
 486     vdwtype          = mdatoms->typeA;
 487
 488     vftab            = kernel_data->table_vdw->data;
 489     vftabscale       = gmx_fjsp_set1_v2r8(kernel_data->table_vdw->scale);
 490
 491     sh_ewald         = gmx_fjsp_set1_v2r8(fr->ic->sh_ewald);
 492     ewtab            = fr->ic->tabq_coul_F;
 493     ewtabscale       = gmx_fjsp_set1_v2r8(fr->ic->tabq_scale);
 494     ewtabhalfspace   = gmx_fjsp_set1_v2r8(0.5/fr->ic->tabq_scale);
 495
 496     /* Avoid stupid compiler warnings */
 497     jnrA = jnrB = 0;
 498     j_coord_offsetA = 0;
 499     j_coord_offsetB = 0;
 500
 501     outeriter        = 0;
 502     inneriter        = 0;
 503
 504     /* Start outer loop over neighborlists */
 505     for(iidx=0; iidx<nri; iidx++)
 506     {
 507         /* Load shift vector for this list */
 508         i_shift_offset   = DIM*shiftidx[iidx];
 509
 510         /* Load limits for loop over neighbors */
 511         j_index_start    = jindex[iidx];
 512         j_index_end      = jindex[iidx+1];
 513
 514         /* Get outer coordinate index */
 515         inr              = iinr[iidx];
 516         i_coord_offset   = DIM*inr;
 517
 518         /* Load i particle coords and add shift vector */
 519         gmx_fjsp_load_shift_and_1rvec_broadcast_v2r8(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 520
 521         fix0             = _fjsp_setzero_v2r8();
 522         fiy0             = _fjsp_setzero_v2r8();
 523         fiz0             = _fjsp_setzero_v2r8();
 524
 525         /* Load parameters for i particles */
 526         iq0              = _fjsp_mul_v2r8(facel,gmx_fjsp_load1_v2r8(charge+inr+0));
 527         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 528
 529         /* Start inner kernel loop */
 530         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 531         {
 532
 533             /* Get j neighbor index, and coordinate index */
 534             jnrA             = jjnr[jidx];
 535             jnrB             = jjnr[jidx+1];
 536             j_coord_offsetA  = DIM*jnrA;
 537             j_coord_offsetB  = DIM*jnrB;
 538
 539             /* load j atom coordinates */
 540             gmx_fjsp_load_1rvec_2ptr_swizzle_v2r8(x+j_coord_offsetA,x+j_coord_offsetB,
 541                                               &jx0,&jy0,&jz0);
 542
 543             /* Calculate displacement vector */
 544             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 545             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 546             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 547
 548             /* Calculate squared distance and things based on it */
 549             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 550
 551             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 552
 553             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 554
 555             /* Load parameters for j particles */
 556             jq0              = gmx_fjsp_load_2real_swizzle_v2r8(charge+jnrA+0,charge+jnrB+0);
 557             vdwjidx0A        = 2*vdwtype[jnrA+0];
 558             vdwjidx0B        = 2*vdwtype[jnrB+0];
 559
 560             /**************************
 561              * CALCULATE INTERACTIONS *
 562              **************************/
 563
 564             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 565
 566             /* Compute parameters for interactions between i and j atoms */
 567             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 568             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 569                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 570
 571             /* Calculate table index by multiplying r with table scale and truncate to integer */
 572             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 573             itab_tmp         = _fjsp_dtox_v2r8(rt);
 574             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 575             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 576             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 577
 578             vfconv.i[0]     *= 8;
 579             vfconv.i[1]     *= 8;
 580
 581             /* EWALD ELECTROSTATICS */
 582
 583             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 584             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 585             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 586             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 587             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 588
 589             gmx_fjsp_load_2pair_swizzle_v2r8(ewtab+ewconv.i[0],ewtab+ewconv.i[1],
 590                                          &ewtabF,&ewtabFn);
 591             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 592             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 593
 594             /* CUBIC SPLINE TABLE DISPERSION */
 595             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 596             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] );
 597             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 598             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 599             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 2 );
 600             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 601             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 602             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 603             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 604
 605             /* CUBIC SPLINE TABLE REPULSION */
 606             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 607             F                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 4 );
 608             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 609             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 610             H                = _fjsp_load_v2r8( vftab + vfconv.i[1] + 6 );
 611             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 612             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 613             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 614             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 615             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 616
 617             fscal            = _fjsp_add_v2r8(felec,fvdw);
 618
 619             /* Update vectorial force */
 620             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 621             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 622             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 623
 624             gmx_fjsp_decrement_fma_1rvec_2ptr_swizzle_v2r8(f+j_coord_offsetA,f+j_coord_offsetB,fscal,dx00,dy00,dz00);
 625
 626             /* Inner loop uses 65 flops */
 627         }
 628
 629         if(jidx<j_index_end)
 630         {
 631
 632             jnrA             = jjnr[jidx];
 633             j_coord_offsetA  = DIM*jnrA;
 634
 635             /* load j atom coordinates */
 636             gmx_fjsp_load_1rvec_1ptr_swizzle_v2r8(x+j_coord_offsetA,
 637                                               &jx0,&jy0,&jz0);
 638
 639             /* Calculate displacement vector */
 640             dx00             = _fjsp_sub_v2r8(ix0,jx0);
 641             dy00             = _fjsp_sub_v2r8(iy0,jy0);
 642             dz00             = _fjsp_sub_v2r8(iz0,jz0);
 643
 644             /* Calculate squared distance and things based on it */
 645             rsq00            = gmx_fjsp_calc_rsq_v2r8(dx00,dy00,dz00);
 646
 647             rinv00           = gmx_fjsp_invsqrt_v2r8(rsq00);
 648
 649             rinvsq00         = _fjsp_mul_v2r8(rinv00,rinv00);
 650
 651             /* Load parameters for j particles */
 652             jq0              = _fjsp_loadl_v2r8(_fjsp_setzero_v2r8(),charge+jnrA+0);
 653             vdwjidx0A        = 2*vdwtype[jnrA+0];
 654
 655             /**************************
 656              * CALCULATE INTERACTIONS *
 657              **************************/
 658
 659             r00              = _fjsp_mul_v2r8(rsq00,rinv00);
 660
 661             /* Compute parameters for interactions between i and j atoms */
 662             qq00             = _fjsp_mul_v2r8(iq0,jq0);
 663             gmx_fjsp_load_2pair_swizzle_v2r8(vdwparam+vdwioffset0+vdwjidx0A,
 664                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 665
 666             /* Calculate table index by multiplying r with table scale and truncate to integer */
 667             rt               = _fjsp_mul_v2r8(r00,vftabscale);
 668             itab_tmp         = _fjsp_dtox_v2r8(rt);
 669             vfeps            = _fjsp_sub_v2r8(rt, _fjsp_xtod_v2r8(itab_tmp));
 670             twovfeps         = _fjsp_add_v2r8(vfeps,vfeps);
 671             _fjsp_store_v2r8(&vfconv.simd,itab_tmp);
 672
 673             vfconv.i[0]     *= 8;
 674             vfconv.i[1]     *= 8;
 675
 676             /* EWALD ELECTROSTATICS */
 677
 678             /* Calculate Ewald table index by multiplying r with scale and truncate to integer */
 679             ewrt             = _fjsp_mul_v2r8(r00,ewtabscale);
 680             itab_tmp         = _fjsp_dtox_v2r8(ewrt);
 681             eweps            = _fjsp_sub_v2r8(ewrt,_fjsp_xtod_v2r8(itab_tmp));
 682             _fjsp_store_v2r8(&ewconv.simd,itab_tmp);
 683
 684             gmx_fjsp_load_1pair_swizzle_v2r8(ewtab+ewconv.i[0],&ewtabF,&ewtabFn);
 685             felec            = _fjsp_madd_v2r8(eweps,ewtabFn,_fjsp_nmsub_v2r8(eweps,ewtabF,ewtabF));
 686             felec            = _fjsp_mul_v2r8(_fjsp_mul_v2r8(qq00,rinv00),_fjsp_sub_v2r8(rinvsq00,felec));
 687
 688             /* CUBIC SPLINE TABLE DISPERSION */
 689             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] );
 690             F                = _fjsp_setzero_v2r8();
 691             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 692             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 2 );
 693             H                = _fjsp_setzero_v2r8();
 694             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 695             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 696             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 697             fvdw6            = _fjsp_mul_v2r8(c6_00,FF);
 698
 699             /* CUBIC SPLINE TABLE REPULSION */
 700             Y                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 4 );
 701             F                = _fjsp_setzero_v2r8();
 702             GMX_FJSP_TRANSPOSE2_V2R8(Y,F);
 703             G                = _fjsp_load_v2r8( vftab + vfconv.i[0] + 6 );
 704             H                = _fjsp_setzero_v2r8();
 705             GMX_FJSP_TRANSPOSE2_V2R8(G,H);
 706             Fp               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(H,vfeps,G),F);
 707             FF               = _fjsp_madd_v2r8(vfeps,_fjsp_madd_v2r8(twovfeps,H,G),Fp);
 708             fvdw12           = _fjsp_mul_v2r8(c12_00,FF);
 709             fvdw             = _fjsp_neg_v2r8(_fjsp_mul_v2r8(_fjsp_add_v2r8(fvdw6,fvdw12),_fjsp_mul_v2r8(vftabscale,rinv00)));
 710
 711             fscal            = _fjsp_add_v2r8(felec,fvdw);
 712
 713             fscal            = _fjsp_unpacklo_v2r8(fscal,_fjsp_setzero_v2r8());
 714
 715             /* Update vectorial force */
 716             fix0             = _fjsp_madd_v2r8(dx00,fscal,fix0);
 717             fiy0             = _fjsp_madd_v2r8(dy00,fscal,fiy0);
 718             fiz0             = _fjsp_madd_v2r8(dz00,fscal,fiz0);
 719
 720             gmx_fjsp_decrement_fma_1rvec_1ptr_swizzle_v2r8(f+j_coord_offsetA,fscal,dx00,dy00,dz00);
 721
 722             /* Inner loop uses 65 flops */
 723         }
 724
 725         /* End of innermost loop */
 726
 727         gmx_fjsp_update_iforce_1atom_swizzle_v2r8(fix0,fiy0,fiz0,
 728                                               f+i_coord_offset,fshift+i_shift_offset);
 729
 730         /* Increment number of inner iterations */
 731         inneriter                  += j_index_end - j_index_start;
 732
 733         /* Outer loop uses 7 flops */
 734     }
 735
 736     /* Increment number of outer iterations */
 737     outeriter        += nri;
 738
 739     /* Update outer/inner flops */
 740
 741     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*65);
 742 }