src/gromacs/mdlib/nbnxn_kernels/simd_2xnn/nbnxn_kernel_simd_2xnn.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 Verlet kernel generator for
  37  * kernel type 2xnn.
  38  */
  39
  40 #ifdef HAVE_CONFIG_H
  41 #include <config.h>
  42 #endif
  43
  44 #include "typedefs.h"
  45
  46 #ifdef GMX_NBNXN_SIMD_2XNN
  47
  48 /* Include the full-width SIMD macros */
  49
  50 #include "gromacs/simd/macros.h"
  51 #include "gromacs/simd/vector_operations.h"
  52
  53 #if !(GMX_SIMD_REAL_WIDTH == 8 || GMX_SIMD_REAL_WIDTH == 16)
  54 #error "unsupported SIMD width"
  55 #endif
  56
  57 #define GMX_SIMD_J_UNROLL_SIZE 2
  58 #include "nbnxn_kernel_simd_2xnn.h"
  59 #include "../nbnxn_kernel_common.h"
  60 #include "gmx_omp_nthreads.h"
  61 #include "types/force_flags.h"
  62 #include "gmx_fatal.h"
  63
  64 /*! \brief Kinds of electrostatic treatments in SIMD Verlet kernels
  65  */
  66 enum {
  67     coulktRF, coulktTAB, coulktTAB_TWIN, coulktEWALD, coulktEWALD_TWIN, coulktNR
  68 };
  69
  70 /*! \brief Kinds of Van der Waals treatments in SIMD Verlet kernels
  71  */
  72 enum {
  73     vdwktLJCUT_COMBGEOM, vdwktLJCUT_COMBLB, vdwktLJCUT_COMBNONE, vdwktLJFORCESWITCH, vdwktLJPOTSWITCH, vdwktLJEWALDCOMBGEOM, vdwktNR
  74 };
  75
  76 /* Declare and define the kernel function pointer lookup tables.
  77  * The minor index of the array goes over both the LJ combination rules,
  78  * which is only supported by plain cut-off, and the LJ switch/PME functions.
  79  */
  80 static p_nbk_func_noener p_nbk_noener[coulktNR][vdwktNR] =
  81 {
  82     {
  83         nbnxn_kernel_ElecRF_VdwLJCombGeom_F_2xnn,
  84         nbnxn_kernel_ElecRF_VdwLJCombLB_F_2xnn,
  85         nbnxn_kernel_ElecRF_VdwLJ_F_2xnn,
  86         nbnxn_kernel_ElecRF_VdwLJFSw_F_2xnn,
  87         nbnxn_kernel_ElecRF_VdwLJPSw_F_2xnn,
  88         nbnxn_kernel_ElecRF_VdwLJEwCombGeom_F_2xnn,
  89     },
  90     {
  91         nbnxn_kernel_ElecQSTab_VdwLJCombGeom_F_2xnn,
  92         nbnxn_kernel_ElecQSTab_VdwLJCombLB_F_2xnn,
  93         nbnxn_kernel_ElecQSTab_VdwLJ_F_2xnn,
  94         nbnxn_kernel_ElecQSTab_VdwLJFSw_F_2xnn,
  95         nbnxn_kernel_ElecQSTab_VdwLJPSw_F_2xnn,
  96         nbnxn_kernel_ElecQSTab_VdwLJEwCombGeom_F_2xnn,
  97     },
  98     {
  99         nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombGeom_F_2xnn,
 100         nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombLB_F_2xnn,
 101         nbnxn_kernel_ElecQSTabTwinCut_VdwLJ_F_2xnn,
 102         nbnxn_kernel_ElecQSTabTwinCut_VdwLJFSw_F_2xnn,
 103         nbnxn_kernel_ElecQSTabTwinCut_VdwLJPSw_F_2xnn,
 104         nbnxn_kernel_ElecQSTabTwinCut_VdwLJEwCombGeom_F_2xnn,
 105     },
 106     {
 107         nbnxn_kernel_ElecEw_VdwLJCombGeom_F_2xnn,
 108         nbnxn_kernel_ElecEw_VdwLJCombLB_F_2xnn,
 109         nbnxn_kernel_ElecEw_VdwLJ_F_2xnn,
 110         nbnxn_kernel_ElecEw_VdwLJFSw_F_2xnn,
 111         nbnxn_kernel_ElecEw_VdwLJPSw_F_2xnn,
 112         nbnxn_kernel_ElecEw_VdwLJEwCombGeom_F_2xnn,
 113     },
 114     {
 115         nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_F_2xnn,
 116         nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_F_2xnn,
 117         nbnxn_kernel_ElecEwTwinCut_VdwLJ_F_2xnn,
 118         nbnxn_kernel_ElecEwTwinCut_VdwLJFSw_F_2xnn,
 119         nbnxn_kernel_ElecEwTwinCut_VdwLJPSw_F_2xnn,
 120         nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_F_2xnn,
 121     },
 122 };
 123
 124 static p_nbk_func_ener p_nbk_ener[coulktNR][vdwktNR] =
 125 {
 126     {
 127         nbnxn_kernel_ElecRF_VdwLJCombGeom_VF_2xnn,
 128         nbnxn_kernel_ElecRF_VdwLJCombLB_VF_2xnn,
 129         nbnxn_kernel_ElecRF_VdwLJ_VF_2xnn,
 130         nbnxn_kernel_ElecRF_VdwLJFSw_VF_2xnn,
 131         nbnxn_kernel_ElecRF_VdwLJPSw_VF_2xnn,
 132         nbnxn_kernel_ElecRF_VdwLJEwCombGeom_VF_2xnn,
 133     },
 134     {
 135         nbnxn_kernel_ElecQSTab_VdwLJCombGeom_VF_2xnn,
 136         nbnxn_kernel_ElecQSTab_VdwLJCombLB_VF_2xnn,
 137         nbnxn_kernel_ElecQSTab_VdwLJ_VF_2xnn,
 138         nbnxn_kernel_ElecQSTab_VdwLJFSw_VF_2xnn,
 139         nbnxn_kernel_ElecQSTab_VdwLJPSw_VF_2xnn,
 140         nbnxn_kernel_ElecQSTab_VdwLJEwCombGeom_VF_2xnn,
 141     },
 142     {
 143         nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombGeom_VF_2xnn,
 144         nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombLB_VF_2xnn,
 145         nbnxn_kernel_ElecQSTabTwinCut_VdwLJ_VF_2xnn,
 146         nbnxn_kernel_ElecQSTabTwinCut_VdwLJFSw_VF_2xnn,
 147         nbnxn_kernel_ElecQSTabTwinCut_VdwLJPSw_VF_2xnn,
 148         nbnxn_kernel_ElecQSTabTwinCut_VdwLJEwCombGeom_VF_2xnn,
 149     },
 150     {
 151         nbnxn_kernel_ElecEw_VdwLJCombGeom_VF_2xnn,
 152         nbnxn_kernel_ElecEw_VdwLJCombLB_VF_2xnn,
 153         nbnxn_kernel_ElecEw_VdwLJ_VF_2xnn,
 154         nbnxn_kernel_ElecEw_VdwLJFSw_VF_2xnn,
 155         nbnxn_kernel_ElecEw_VdwLJPSw_VF_2xnn,
 156         nbnxn_kernel_ElecEw_VdwLJEwCombGeom_VF_2xnn,
 157     },
 158     {
 159         nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_VF_2xnn,
 160         nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_VF_2xnn,
 161         nbnxn_kernel_ElecEwTwinCut_VdwLJ_VF_2xnn,
 162         nbnxn_kernel_ElecEwTwinCut_VdwLJFSw_VF_2xnn,
 163         nbnxn_kernel_ElecEwTwinCut_VdwLJPSw_VF_2xnn,
 164         nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_VF_2xnn,
 165     },
 166 };
 167
 168 static p_nbk_func_ener p_nbk_energrp[coulktNR][vdwktNR] =
 169 {
 170     {
 171         nbnxn_kernel_ElecRF_VdwLJCombGeom_VgrpF_2xnn,
 172         nbnxn_kernel_ElecRF_VdwLJCombLB_VgrpF_2xnn,
 173         nbnxn_kernel_ElecRF_VdwLJ_VgrpF_2xnn,
 174         nbnxn_kernel_ElecRF_VdwLJFSw_VgrpF_2xnn,
 175         nbnxn_kernel_ElecRF_VdwLJPSw_VgrpF_2xnn,
 176         nbnxn_kernel_ElecRF_VdwLJEwCombGeom_VgrpF_2xnn,
 177     },
 178     {
 179         nbnxn_kernel_ElecQSTab_VdwLJCombGeom_VgrpF_2xnn,
 180         nbnxn_kernel_ElecQSTab_VdwLJCombLB_VgrpF_2xnn,
 181         nbnxn_kernel_ElecQSTab_VdwLJ_VgrpF_2xnn,
 182         nbnxn_kernel_ElecQSTab_VdwLJFSw_VgrpF_2xnn,
 183         nbnxn_kernel_ElecQSTab_VdwLJPSw_VgrpF_2xnn,
 184         nbnxn_kernel_ElecQSTab_VdwLJEwCombGeom_VgrpF_2xnn,
 185     },
 186     {
 187         nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombGeom_VgrpF_2xnn,
 188         nbnxn_kernel_ElecQSTabTwinCut_VdwLJCombLB_VgrpF_2xnn,
 189         nbnxn_kernel_ElecQSTabTwinCut_VdwLJ_VgrpF_2xnn,
 190         nbnxn_kernel_ElecQSTabTwinCut_VdwLJFSw_VgrpF_2xnn,
 191         nbnxn_kernel_ElecQSTabTwinCut_VdwLJPSw_VgrpF_2xnn,
 192         nbnxn_kernel_ElecQSTabTwinCut_VdwLJEwCombGeom_VgrpF_2xnn,
 193     },
 194     {
 195         nbnxn_kernel_ElecEw_VdwLJCombGeom_VgrpF_2xnn,
 196         nbnxn_kernel_ElecEw_VdwLJCombLB_VgrpF_2xnn,
 197         nbnxn_kernel_ElecEw_VdwLJ_VgrpF_2xnn,
 198         nbnxn_kernel_ElecEw_VdwLJFSw_VgrpF_2xnn,
 199         nbnxn_kernel_ElecEw_VdwLJPSw_VgrpF_2xnn,
 200         nbnxn_kernel_ElecEw_VdwLJEwCombGeom_VgrpF_2xnn,
 201     },
 202     {
 203         nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_VgrpF_2xnn,
 204         nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_VgrpF_2xnn,
 205         nbnxn_kernel_ElecEwTwinCut_VdwLJ_VgrpF_2xnn,
 206         nbnxn_kernel_ElecEwTwinCut_VdwLJFSw_VgrpF_2xnn,
 207         nbnxn_kernel_ElecEwTwinCut_VdwLJPSw_VgrpF_2xnn,
 208         nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_VgrpF_2xnn,
 209     },
 210 };
 211
 212
 213 static void
 214 reduce_group_energies(int ng, int ng_2log,
 215                       const real *VSvdw, const real *VSc,
 216                       real *Vvdw, real *Vc)
 217 {
 218     const int unrollj      = GMX_SIMD_REAL_WIDTH/GMX_SIMD_J_UNROLL_SIZE;
 219     const int unrollj_half = unrollj/2;
 220     int       ng_p2, i, j, j0, j1, c, s;
 221
 222     ng_p2 = (1<<ng_2log);
 223
 224     /* The size of the x86 SIMD energy group buffer array is:
 225      * ng*ng*ng_p2*unrollj_half*simd_width
 226      */
 227     for (i = 0; i < ng; i++)
 228     {
 229         for (j = 0; j < ng; j++)
 230         {
 231             Vvdw[i*ng+j] = 0;
 232             Vc[i*ng+j]   = 0;
 233         }
 234
 235         for (j1 = 0; j1 < ng; j1++)
 236         {
 237             for (j0 = 0; j0 < ng; j0++)
 238             {
 239                 c = ((i*ng + j1)*ng_p2 + j0)*unrollj_half*unrollj;
 240                 for (s = 0; s < unrollj_half; s++)
 241                 {
 242                     Vvdw[i*ng+j0] += VSvdw[c+0];
 243                     Vvdw[i*ng+j1] += VSvdw[c+1];
 244                     Vc  [i*ng+j0] += VSc  [c+0];
 245                     Vc  [i*ng+j1] += VSc  [c+1];
 246                     c             += unrollj + 2;
 247                 }
 248             }
 249         }
 250     }
 251 }
 252
 253 #else /* GMX_NBNXN_SIMD_2XNN */
 254
 255 #include "gmx_fatal.h"
 256
 257 #endif /* GMX_NBNXN_SIMD_2XNN */
 258
 259 void
 260 nbnxn_kernel_simd_2xnn(nbnxn_pairlist_set_t      gmx_unused *nbl_list,
 261                        const nbnxn_atomdata_t    gmx_unused *nbat,
 262                        const interaction_const_t gmx_unused *ic,
 263                        int                       gmx_unused  ewald_excl,
 264                        rvec                      gmx_unused *shift_vec,
 265                        int                       gmx_unused  force_flags,
 266                        int                       gmx_unused  clearF,
 267                        real                      gmx_unused *fshift,
 268                        real                      gmx_unused *Vc,
 269                        real                      gmx_unused *Vvdw)
 270 #ifdef GMX_NBNXN_SIMD_2XNN
 271 {
 272     int                nnbl;
 273     nbnxn_pairlist_t **nbl;
 274     int                coulkt, vdwkt = 0;
 275     int                nb;
 276
 277     nnbl = nbl_list->nnbl;
 278     nbl  = nbl_list->nbl;
 279
 280     if (EEL_RF(ic->eeltype) || ic->eeltype == eelCUT)
 281     {
 282         coulkt = coulktRF;
 283     }
 284     else
 285     {
 286         if (ewald_excl == ewaldexclTable)
 287         {
 288             if (ic->rcoulomb == ic->rvdw)
 289             {
 290                 coulkt = coulktTAB;
 291             }
 292             else
 293             {
 294                 coulkt = coulktTAB_TWIN;
 295             }
 296         }
 297         else
 298         {
 299             if (ic->rcoulomb == ic->rvdw)
 300             {
 301                 coulkt = coulktEWALD;
 302             }
 303             else
 304             {
 305                 coulkt = coulktEWALD_TWIN;
 306             }
 307         }
 308     }
 309
 310     if (ic->vdwtype == evdwCUT)
 311     {
 312         switch (ic->vdw_modifier)
 313         {
 314             case eintmodNONE:
 315             case eintmodPOTSHIFT:
 316                 switch (nbat->comb_rule)
 317                 {
 318                     case ljcrGEOM: vdwkt = vdwktLJCUT_COMBGEOM; break;
 319                     case ljcrLB:   vdwkt = vdwktLJCUT_COMBLB;   break;
 320                     case ljcrNONE: vdwkt = vdwktLJCUT_COMBNONE; break;
 321                     default:       gmx_incons("Unknown combination rule");
 322                 }
 323                 break;
 324             case eintmodFORCESWITCH:
 325                 vdwkt = vdwktLJFORCESWITCH;
 326                 break;
 327             case eintmodPOTSWITCH:
 328                 vdwkt = vdwktLJPOTSWITCH;
 329                 break;
 330             default:
 331                 gmx_incons("Unsupported VdW interaction modifier");
 332         }
 333     }
 334     else if (ic->vdwtype == evdwPME)
 335     {
 336         if (ic->ljpme_comb_rule == eljpmeLB)
 337         {
 338             gmx_incons("The nbnxn SIMD kernels don't suport LJ-PME with LB");
 339         }
 340         vdwkt = vdwktLJEWALDCOMBGEOM;
 341     }
 342     else
 343     {
 344         gmx_incons("Unsupported VdW interaction type");
 345     }
 346
 347 #pragma omp parallel for schedule(static) num_threads(gmx_omp_nthreads_get(emntNonbonded))
 348     for (nb = 0; nb < nnbl; nb++)
 349     {
 350         nbnxn_atomdata_output_t *out;
 351         real                    *fshift_p;
 352
 353         out = &nbat->out[nb];
 354
 355         if (clearF == enbvClearFYes)
 356         {
 357             clear_f(nbat, nb, out->f);
 358         }
 359
 360         if ((force_flags & GMX_FORCE_VIRIAL) && nnbl == 1)
 361         {
 362             fshift_p = fshift;
 363         }
 364         else
 365         {
 366             fshift_p = out->fshift;
 367
 368             if (clearF == enbvClearFYes)
 369             {
 370                 clear_fshift(fshift_p);
 371             }
 372         }
 373
 374         if (!(force_flags & GMX_FORCE_ENERGY))
 375         {
 376             /* Don't calculate energies */
 377             p_nbk_noener[coulkt][vdwkt](nbl[nb], nbat,
 378                                         ic,
 379                                         shift_vec,
 380                                         out->f,
 381                                         fshift_p);
 382         }
 383         else if (out->nV == 1)
 384         {
 385             /* No energy groups */
 386             out->Vvdw[0] = 0;
 387             out->Vc[0]   = 0;
 388
 389             p_nbk_ener[coulkt][vdwkt](nbl[nb], nbat,
 390                                       ic,
 391                                       shift_vec,
 392                                       out->f,
 393                                       fshift_p,
 394                                       out->Vvdw,
 395                                       out->Vc);
 396         }
 397         else
 398         {
 399             /* Calculate energy group contributions */
 400             int i;
 401
 402             for (i = 0; i < out->nVS; i++)
 403             {
 404                 out->VSvdw[i] = 0;
 405             }
 406             for (i = 0; i < out->nVS; i++)
 407             {
 408                 out->VSc[i] = 0;
 409             }
 410
 411             p_nbk_energrp[coulkt][vdwkt](nbl[nb], nbat,
 412                                          ic,
 413                                          shift_vec,
 414                                          out->f,
 415                                          fshift_p,
 416                                          out->VSvdw,
 417                                          out->VSc);
 418
 419             reduce_group_energies(nbat->nenergrp, nbat->neg_2log,
 420                                   out->VSvdw, out->VSc,
 421                                   out->Vvdw, out->Vc);
 422         }
 423     }
 424
 425     if (force_flags & GMX_FORCE_ENERGY)
 426     {
 427         reduce_energies_over_lists(nbat, nnbl, Vvdw, Vc);
 428     }
 429 }
 430 #else
 431 {
 432     gmx_incons("nbnxn_kernel_simd_2xnn called when such kernels "
 433                " are not enabled.");
 434 }
 435 #endif
 436 #undef GMX_SIMD_J_UNROLL_SIZE