2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2020,2021, 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.
35 #ifndef GMX_GPU_UTILS_DEVICE_STREAM_H
36 #define GMX_GPU_UTILS_DEVICE_STREAM_H
38 /*! \libinternal \file
40 * \brief Declarations for DeviceStream class.
42 * \author Artem Zhmurov <zhmurov@gmail.com>
43 * \author Mark Abraham <mark.j.abraham@gmail.com>
45 * \ingroup module_gpu_utils
54 # include <cuda_runtime.h>
57 # include "gromacs/gpu_utils/gmxopencl.h"
59 # include "gromacs/gpu_utils/gmxsycl.h"
62 #include "gromacs/utility/classhelpers.h"
64 struct DeviceInformation;
67 //! Enumeration describing the priority with which a stream operates.
68 enum class DeviceStreamPriority : int
70 //! High-priority stream
72 //! Normal-priority stream
74 //! Conventional termination of the enumeration
78 /*! \libinternal \brief Declaration of platform-agnostic device stream/queue.
80 * The command stream (or command queue) is a sequence of operations that are executed
81 * in they order they were issued. Several streams may co-exist to represent concurrency.
82 * This class declares the interfaces, that are exposed to platform-agnostic code and
83 * it should be implemented for each compute architecture (e.g. CUDA and OpenCL).
85 * Destruction of the \p DeviceStream calls the destructor of the underlying low-level
86 * stream/queue, hence should only be called when the stream is no longer needed. To
87 * prevent accidental stream destruction, while copying or moving a \p DeviceStream
88 * object, copy and move constructors and copy and move assignments are not allowed
89 * and the \p DeviceStream object should be passed as a pointer or constant reference.
95 /*! \brief Construct and init.
97 * \param[in] deviceContext Device context (only used in OpenCL and SYCL).
98 * \param[in] priority Stream priority: high or normal (only used in CUDA).
99 * \param[in] useTiming If the timing should be enabled (only used in OpenCL and SYCL).
101 DeviceStream(const DeviceContext& deviceContext, DeviceStreamPriority priority, bool useTiming);
106 /*! \brief Check if the underlying stream is valid.
108 * \returns Whether the stream is valid (false in CPU-only builds).
110 bool isValid() const;
112 //! Synchronize the stream
113 void synchronize() const;
118 cudaStream_t stream() const;
121 cudaStream_t stream_ = nullptr;
125 * Getter for the underlying \c cl::sycl:queue object.
127 * Returns a copy instead of const-reference, because it's impossible to submit to or wait
128 * on a \c const cl::sycl::queue. SYCL standard guarantees that operating on copy is
129 * equivalent to operating on the original queue.
131 * \throws std::bad_optional_access if the stream is not valid.
133 * \returns A copy of the internal \c cl::sycl:queue.
135 cl::sycl::queue stream() const { return cl::sycl::queue(stream_); }
136 //! Getter. Can throw std::bad_optional_access if the stream is not valid.
137 cl::sycl::queue& stream() { return stream_; }
138 //! Synchronize the stream. Non-const version of \c ::synchronize() for SYCL that does not do unnecessary copying.
142 cl::sycl::queue stream_;
143 #elif GMX_GPU_OPENCL || defined DOXYGEN
146 cl_command_queue stream() const;
149 cl_command_queue stream_ = nullptr;
153 GMX_DISALLOW_COPY_MOVE_AND_ASSIGN(DeviceStream);
156 #endif // GMX_GPU_UTILS_DEVICE_STREAM_H