src/gromacs/timing/gpu_timing.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2012,2014,2015,2016,2017 by the GROMACS development team.
   5  * Copyright (c) 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 /*! \libinternal \file
  37  *  \brief Declares data types for GPU timing
  38  *
  39  *  \author Szilard Pall <pall.szilard@gmail.com>
  40  *  \author Mark Abraham <mark.j.abraham@gmail.com>
  41  *  \inlibraryapi
  42  */
  43
  44 #ifndef GMX_TIMING_GPU_TIMING_H
  45 #define GMX_TIMING_GPU_TIMING_H
  46
  47 /*! \internal \brief GPU kernel time and call count. */
  48 struct gmx_kernel_timing_data_t
  49 {
  50     double t; /**< Accumulated lapsed time */
  51     int    c; /**< Number of calls corresponding to the elapsed time */
  52 };
  53
  54 /*! \internal \brief
  55  * PME GPU stages timing events indices, corresponding to the string in PMEStageNames in wallcycle.cpp.
  56  */
  57 enum
  58 {
  59     gtPME_SPLINE = 0,
  60     gtPME_SPREAD,
  61     gtPME_SPLINEANDSPREAD,
  62     gtPME_FFT_R2C,
  63     gtPME_SOLVE,
  64     gtPME_FFT_C2R,
  65     gtPME_GATHER,
  66     gtPME_EVENT_COUNT /* not a stage ID but a static array size */
  67 };
  68
  69 /*! \internal \brief GPU timings for PME. */
  70 struct gmx_wallclock_gpu_pme_t
  71 {
  72     /* A separate PME structure to avoid refactoring the NB code for gmx_wallclock_gpu_t later
  73      * TODO: devise a better GPU timing data structuring.
  74      */
  75     /*! \brief Array of PME GPU timing data. */
  76     gmx_kernel_timing_data_t timing[gtPME_EVENT_COUNT];
  77 };
  78
  79 /*! \internal \brief GPU NB timings for kernels and H2d/D2H transfers. */
  80 struct gmx_wallclock_gpu_nbnxn_t
  81 {
  82     gmx_kernel_timing_data_t ktime[2][2]; /**< table containing the timings of the four
  83                                                    versions of the nonbonded kernels: force-only,
  84                                                    force+energy, force+pruning, and force+energy+pruning */
  85     gmx_kernel_timing_data_t pruneTime; /**< table containing the timings of the 1st pass prune-only kernels */
  86     gmx_kernel_timing_data_t dynamicPruneTime; /**< table containing the timings of dynamic prune-only kernels */
  87     double                   nb_h2d_t; /**< host to device transfer time in nb calculation  */
  88     double                   nb_d2h_t; /**< device to host transfer time in nb calculation */
  89     int                      nb_c;     /**< total call count of the nonbonded gpu operations */
  90     double                   pl_h2d_t; /**< pair search step host to device transfer time */
  91     int                      pl_h2d_c; /**< pair search step  host to device transfer call count */
  92 };
  93
  94 #endif