src/gromacs/ewald/pme_gpu_settings.h

   1 /*
   2  * This file is part of the GROMACS molecular simulation package.
   3  *
   4  * Copyright (c) 2020, 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 /*! \libinternal \file
  37  * \brief Defines the PME GPU settings data structures.
  38  * \todo Some renaming/refactoring, which does not impair the performance:
  39  * -- PmeGpuSettings -> PmeGpuTasks
  40  *
  41  * \author Aleksei Iupinov <a.yupinov@gmail.com>
  42  * \author Mark Abraham <mark.j.abraham@gmail.com>
  43  * \ingroup module_ewald
  44  */
  45
  46 #ifndef GMX_EWALD_PME_GPU_SETTINGS_H
  47 #define GMX_EWALD_PME_GPU_SETTINGS_H
  48
  49 #include "gromacs/gpu_utils/gpu_utils.h" // for GpuApiCallBehavior
  50
  51 enum class ThreadsPerAtom : int;
  52
  53 /*! \internal \brief
  54  * The PME GPU settings structure, included in the main PME GPU structure by value.
  55  */
  56 struct PmeGpuSettings
  57 {
  58     /* Permanent settings set on initialization */
  59     /*! \brief A boolean which tells if the solving is performed on GPU. Currently always true */
  60     bool performGPUSolve;
  61     /*! \brief A boolean which tells if the gathering is performed on GPU. Currently always true */
  62     bool performGPUGather;
  63     /*! \brief A boolean which tells if the FFT is performed on GPU. Currently true for a single MPI rank. */
  64     bool performGPUFFT;
  65     /*! \brief A convenience boolean which tells if PME decomposition is used. */
  66     bool useDecomposition;
  67     /*! \brief True if PME forces are reduced on-GPU, false if reduction is done on the CPU;
  68      *  in the former case transfer does not need to happen.
  69      *
  70      *  Note that this flag may change per-step.
  71      */
  72     bool useGpuForceReduction;
  73
  74     /*! \brief A boolean which tells if any PME GPU stage should copy all of its outputs to the
  75      * host. Only intended to be used by the test framework.
  76      */
  77     bool copyAllOutputs;
  78     /*! \brief An enum which tells whether most PME GPU D2H/H2D data transfers should be synchronous. */
  79     GpuApiCallBehavior transferKind;
  80     /*! \brief
  81      *  Controls whether we use order (i.e. 4) threads per atom for the GPU
  82      *  or order*order (i.e. 16) threads per atom.
  83      *
  84      *  Currently ThreadsPerAtom::Order is only supported by CUDA.
  85      */
  86     ThreadsPerAtom threadsPerAtom;
  87     /*! \brief
  88      * Currently only supported by CUDA.
  89      * Controls if we should recalculate the splines in the gather or
  90      * save the values in the spread and reload in the gather.
  91      */
  92     bool recalculateSplines;
  93 };
  94
  95 #endif