2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2019,2020, 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.
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.
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.
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.
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.
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.
37 * \brief Implements GPU stream manager.
39 * \author Mark Abraham <mark.j.abraham@gmail.com>
40 * \author Artem Zhmurov <zhmurov@gmail.com>
42 * \ingroup module_gpu_utils
46 #include "device_stream_manager.h"
48 #include "gromacs/gpu_utils/device_context.h"
49 #include "gromacs/gpu_utils/device_stream.h"
50 #include "gromacs/mdtypes/simulation_workload.h"
51 #include "gromacs/utility/enumerationhelpers.h"
52 #include "gromacs/utility/exceptions.h"
53 #include "gromacs/utility/gmxassert.h"
54 #include "gromacs/utility/stringutil.h"
60 * \brief Impl class to manages the lifetime of the GPU streams.
62 * If supported by the GPU API, the available runtime and the
63 * indicated device, some streams will be configured at high
64 * priority. Otherwise, all streams will share the default priority
65 * appropriate to the situation.
67 class DeviceStreamManager::Impl
70 /*! \brief Constructor.
72 * \throws InternalError If any of the required resources could not be initialized.
74 Impl(const DeviceInformation& deviceInfo,
75 bool havePpDomainDecomposition,
76 SimulationWorkload simulationWork,
81 DeviceContext context_;
82 //! GPU command streams.
83 EnumerationArray<DeviceStreamType, DeviceStream> streams_;
86 // DeviceStreamManager::Impl
87 DeviceStreamManager::Impl::Impl(const DeviceInformation& deviceInfo,
88 const bool havePpDomainDecomposition,
89 const SimulationWorkload simulationWork,
90 const bool useTiming) :
95 streams_[DeviceStreamType::NonBondedLocal].init(context_, DeviceStreamPriority::Normal, useTiming);
97 if (simulationWork.useGpuPme)
99 /* Creating a PME GPU stream:
100 * - default high priority with CUDA
101 * - no priorities implemented yet with OpenCL; see #2532
103 streams_[DeviceStreamType::Pme].init(context_, DeviceStreamPriority::High, useTiming);
106 if (havePpDomainDecomposition)
108 streams_[DeviceStreamType::NonBondedNonLocal].init(context_, DeviceStreamPriority::High,
111 // Update stream is used both for coordinates transfers and for GPU update/constraints
112 if (simulationWork.useGpuPme || simulationWork.useGpuUpdate || simulationWork.useGpuBufferOps)
114 streams_[DeviceStreamType::UpdateAndConstraints].init(
115 context_, DeviceStreamPriority::Normal, useTiming);
117 if (simulationWork.useGpuPmePpCommunication)
119 streams_[DeviceStreamType::PmePpTransfer].init(context_, DeviceStreamPriority::Normal, useTiming);
122 GMX_CATCH_ALL_AND_EXIT_WITH_FATAL_ERROR
125 DeviceStreamManager::Impl::~Impl() = default;
127 // DeviceStreamManager
128 DeviceStreamManager::DeviceStreamManager(const DeviceInformation& deviceInfo,
129 const bool havePpDomainDecomposition,
130 const SimulationWorkload simulationWork,
131 const bool useTiming) :
132 impl_(new Impl(deviceInfo, havePpDomainDecomposition, simulationWork, useTiming))
136 DeviceStreamManager::~DeviceStreamManager() = default;
138 const DeviceInformation& DeviceStreamManager::deviceInfo() const
140 return impl_->context_.deviceInfo();
143 const DeviceContext& DeviceStreamManager::context() const
145 return impl_->context_;
148 const DeviceStream& DeviceStreamManager::stream(DeviceStreamType streamToGet) const
150 return impl_->streams_[streamToGet];
153 const DeviceStream& DeviceStreamManager::bondedStream(bool hasPPDomainDecomposition) const
155 if (hasPPDomainDecomposition)
157 GMX_RELEASE_ASSERT(stream(DeviceStreamType::NonBondedNonLocal).isValid(),
158 "GPU non-bonded non-local stream should be valid in order to use GPU "
159 "version of bonded forces with domain decomposition.");
160 return stream(DeviceStreamType::NonBondedNonLocal);
164 GMX_RELEASE_ASSERT(stream(DeviceStreamType::NonBondedLocal).isValid(),
165 "GPU non-bonded local stream should be valid in order to use GPU "
166 "version of bonded forces without domain decomposition.");
167 return stream(DeviceStreamType::NonBondedLocal);
171 bool DeviceStreamManager::streamIsValid(DeviceStreamType streamToCheck) const
173 return impl_->streams_[streamToCheck].isValid();