src/gromacs/nbnxm/nbnxm_gpu_data_mgmt.cpp

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2013,2014,2015,2016 by the GROMACS development team.
   5  * Copyright (c) 2017,2018,2019,2020, by the GROMACS development team, led by
   6  * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
   7  * and including many others, as listed in the AUTHORS file in the
   8  * top-level source directory and at http://www.gromacs.org.
   9  *
  10  * GROMACS is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU Lesser General Public License
  12  * as published by the Free Software Foundation; either version 2.1
  13  * of the License, or (at your option) any later version.
  14  *
  15  * GROMACS is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * Lesser General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU Lesser General Public
  21  * License along with GROMACS; if not, see
  22  * http://www.gnu.org/licenses, or write to the Free Software Foundation,
  23  * Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA.
  24  *
  25  * If you want to redistribute modifications to GROMACS, please
  26  * consider that scientific software is very special. Version
  27  * control is crucial - bugs must be traceable. We will be happy to
  28  * consider code for inclusion in the official distribution, but
  29  * derived work must not be called official GROMACS. Details are found
  30  * in the README & COPYING files - if they are missing, get the
  31  * official version at http://www.gromacs.org.
  32  *
  33  * To help us fund GROMACS development, we humbly ask that you cite
  34  * the research papers on the package. Check out http://www.gromacs.org.
  35  */
  36 /*! \internal \file
  37  *  \brief Define common implementation of nbnxm_gpu_data_mgmt.h
  38  *
  39  *  \author Anca Hamuraru <anca@streamcomputing.eu>
  40  *  \author Dimitrios Karkoulis <dimitris.karkoulis@gmail.com>
  41  *  \author Teemu Virolainen <teemu@streamcomputing.eu>
  42  *  \author Szilárd Páll <pall.szilard@gmail.com>
  43  *  \author Artem Zhmurov <zhmurov@gmail.com>
  44  *
  45  *  \ingroup module_nbnxm
  46  */
  47 #include "gmxpre.h"
  48
  49 #include "config.h"
  50
  51 #if GMX_GPU_CUDA
  52 #    include "cuda/nbnxm_cuda_types.h"
  53 #endif
  54
  55 #if GMX_GPU_OPENCL
  56 #    include "opencl/nbnxm_ocl_types.h"
  57 #endif
  58
  59 #include "nbnxm_gpu_data_mgmt.h"
  60
  61 #include "gromacs/timing/gpu_timing.h"
  62
  63 #include "nbnxm_gpu.h"
  64
  65 namespace Nbnxm
  66 {
  67
  68 void init_ewald_coulomb_force_table(const EwaldCorrectionTables& tables,
  69                                     NBParamGpu*                  nbp,
  70                                     const DeviceContext&         deviceContext)
  71 {
  72     if (!nbp->coulomb_tab)
  73     {
  74         destroyParamLookupTable(&nbp->coulomb_tab, nbp->coulomb_tab_texobj);
  75     }
  76
  77     nbp->coulomb_tab_scale = tables.scale;
  78     initParamLookupTable(&nbp->coulomb_tab, &nbp->coulomb_tab_texobj, tables.tableF.data(),
  79                          tables.tableF.size(), deviceContext);
  80 }
  81
  82 int nbnxn_gpu_pick_ewald_kernel_type(const interaction_const_t& ic)
  83 {
  84     bool bTwinCut = (ic.rcoulomb != ic.rvdw);
  85     bool bUseAnalyticalEwald, bForceAnalyticalEwald, bForceTabulatedEwald;
  86     int  kernel_type;
  87
  88     /* Benchmarking/development environment variables to force the use of
  89        analytical or tabulated Ewald kernel. */
  90     bForceAnalyticalEwald = (getenv("GMX_GPU_NB_ANA_EWALD") != nullptr);
  91     bForceTabulatedEwald  = (getenv("GMX_GPU_NB_TAB_EWALD") != nullptr);
  92
  93     if (bForceAnalyticalEwald && bForceTabulatedEwald)
  94     {
  95         gmx_incons(
  96                 "Both analytical and tabulated Ewald GPU non-bonded kernels "
  97                 "requested through environment variables.");
  98     }
  99
 100     /* By default, use analytical Ewald
 101      * TODO: tabulated does not work in OpenCL, it needs fixing, see init_nbparam() in nbnxn_ocl_data_mgmt.cpp
 102      *
 103      */
 104     bUseAnalyticalEwald = true;
 105     if (bForceAnalyticalEwald)
 106     {
 107         if (debug)
 108         {
 109             fprintf(debug, "Using analytical Ewald GPU kernels\n");
 110         }
 111     }
 112     else if (bForceTabulatedEwald)
 113     {
 114         bUseAnalyticalEwald = false;
 115
 116         if (debug)
 117         {
 118             fprintf(debug, "Using tabulated Ewald GPU kernels\n");
 119         }
 120     }
 121
 122     /* Use twin cut-off kernels if requested by bTwinCut or the env. var.
 123        forces it (use it for debugging/benchmarking only). */
 124     if (!bTwinCut && (getenv("GMX_GPU_NB_EWALD_TWINCUT") == nullptr))
 125     {
 126         kernel_type = bUseAnalyticalEwald ? eelTypeEWALD_ANA : eelTypeEWALD_TAB;
 127     }
 128     else
 129     {
 130         kernel_type = bUseAnalyticalEwald ? eelTypeEWALD_ANA_TWIN : eelTypeEWALD_TAB_TWIN;
 131     }
 132
 133     return kernel_type;
 134 }
 135
 136 void set_cutoff_parameters(NBParamGpu* nbp, const interaction_const_t* ic, const PairlistParams& listParams)
 137 {
 138     nbp->ewald_beta        = ic->ewaldcoeff_q;
 139     nbp->sh_ewald          = ic->sh_ewald;
 140     nbp->epsfac            = ic->epsfac;
 141     nbp->two_k_rf          = 2.0 * ic->k_rf;
 142     nbp->c_rf              = ic->c_rf;
 143     nbp->rvdw_sq           = ic->rvdw * ic->rvdw;
 144     nbp->rcoulomb_sq       = ic->rcoulomb * ic->rcoulomb;
 145     nbp->rlistOuter_sq     = listParams.rlistOuter * listParams.rlistOuter;
 146     nbp->rlistInner_sq     = listParams.rlistInner * listParams.rlistInner;
 147     nbp->useDynamicPruning = listParams.useDynamicPruning;
 148
 149     nbp->sh_lj_ewald   = ic->sh_lj_ewald;
 150     nbp->ewaldcoeff_lj = ic->ewaldcoeff_lj;
 151
 152     nbp->rvdw_switch      = ic->rvdw_switch;
 153     nbp->dispersion_shift = ic->dispersion_shift;
 154     nbp->repulsion_shift  = ic->repulsion_shift;
 155     nbp->vdw_switch       = ic->vdw_switch;
 156 }
 157
 158 void init_plist(gpu_plist* pl)
 159 {
 160     /* initialize to nullptr pointers to data that is not allocated here and will
 161        need reallocation in nbnxn_gpu_init_pairlist */
 162     pl->sci   = nullptr;
 163     pl->cj4   = nullptr;
 164     pl->imask = nullptr;
 165     pl->excl  = nullptr;
 166
 167     /* size -1 indicates that the respective array hasn't been initialized yet */
 168     pl->na_c          = -1;
 169     pl->nsci          = -1;
 170     pl->sci_nalloc    = -1;
 171     pl->ncj4          = -1;
 172     pl->cj4_nalloc    = -1;
 173     pl->nimask        = -1;
 174     pl->imask_nalloc  = -1;
 175     pl->nexcl         = -1;
 176     pl->excl_nalloc   = -1;
 177     pl->haveFreshList = false;
 178 }
 179
 180 void init_timings(gmx_wallclock_gpu_nbnxn_t* t)
 181 {
 182     int i, j;
 183
 184     t->nb_h2d_t = 0.0;
 185     t->nb_d2h_t = 0.0;
 186     t->nb_c     = 0;
 187     t->pl_h2d_t = 0.0;
 188     t->pl_h2d_c = 0;
 189     for (i = 0; i < 2; i++)
 190     {
 191         for (j = 0; j < 2; j++)
 192         {
 193             t->ktime[i][j].t = 0.0;
 194             t->ktime[i][j].c = 0;
 195         }
 196     }
 197     t->pruneTime.c        = 0;
 198     t->pruneTime.t        = 0.0;
 199     t->dynamicPruneTime.c = 0;
 200     t->dynamicPruneTime.t = 0.0;
 201 }
 202
 203 } // namespace Nbnxm