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 Declaration of low-level functions and fields of GPU state propagator object.
39 * \author Artem Zhmurov <zhmurov@gmail.com>
41 * \ingroup module_mdtypes
43 #ifndef GMX_MDTYPES_STATE_PROPAGATOR_DATA_GPU_IMPL_H
44 #define GMX_MDTYPES_STATE_PROPAGATOR_DATA_GPU_IMPL_H
50 #include "gromacs/gpu_utils/devicebuffer.h"
52 # include "gromacs/gpu_utils/gpueventsynchronizer.cuh"
54 # include "gromacs/gpu_utils/gpueventsynchronizer_ocl.h"
56 # include "gromacs/gpu_utils/gpueventsynchronizer_sycl.h"
58 #include "gromacs/math/vectypes.h"
59 #include "gromacs/mdtypes/state_propagator_data_gpu.h"
60 #include "gromacs/utility/classhelpers.h"
61 #include "gromacs/utility/enumerationhelpers.h"
68 class StatePropagatorDataGpu::Impl
74 /*! \brief Constructor
76 * The buffers are reallocated only at the reinit call, the padding is
77 * used there for the coordinates buffer. It is needed for PME and added at
78 * the end of the buffer. It is assumed that if the rank has PME duties on the
79 * GPU, all coordinates are copied to the GPU and hence, for this rank, the
80 * coordinates buffer is not split into local and non-local ranges. For other
81 * ranks, the padding size is zero. This works because only one rank ever does
82 * PME work on the GPU, and if that rank also does PP work that is the only
83 * rank. So all coordinates are always transferred.
85 * In OpenCL, only pmeStream is used since it is the only stream created in
86 * PME context. The local and non-local streams are only needed when buffer
87 * ops are offloaded. This feature is currently not available in OpenCL and
88 * hence these streams are not set in these builds.
90 * \param[in] deviceStreamManager Object that owns the DeviceContext and DeviceStreams.
91 * \param[in] transferKind H2D/D2H transfer call behavior (synchronous or not).
92 * \param[in] allocationBlockSizeDivisor Determines the padding size for coordinates buffer.
93 * \param[in] wcycle Wall cycle counter data.
95 Impl(const DeviceStreamManager& deviceStreamManager,
96 GpuApiCallBehavior transferKind,
97 int allocationBlockSizeDivisor,
98 gmx_wallcycle* wcycle);
100 /*! \brief Constructor to use in PME-only rank and in tests.
102 * This constructor should be used if only a coordinate device buffer should be managed
103 * using a single stream. Any operation on force or velocity buffer as well as copy of
104 * non-local coordinates will exit with assertion failure. Note, that the pmeStream can
105 * not be a nullptr and the constructor will exit with an assertion failure.
107 * \todo Currently, unsupported copy operations are blocked by assertion that the stream
108 * not nullptr. This should be improved.
110 * \param[in] pmeStream Device PME stream, nullptr is not allowed.
111 * \param[in] deviceContext Device context, nullptr allowed for non-OpenCL builds.
112 * \param[in] transferKind H2D/D2H transfer call behavior (synchronous or not).
113 * \param[in] allocationBlockSizeDivisor Determines the padding size for coordinates buffer.
114 * \param[in] wcycle Wall cycle counter data.
116 Impl(const DeviceStream* pmeStream,
117 const DeviceContext& deviceContext,
118 GpuApiCallBehavior transferKind,
119 int allocationBlockSizeDivisor,
120 gmx_wallcycle* wcycle);
125 /*! \brief Set the ranges for local and non-local atoms and reallocates buffers.
127 * Reallocates coordinate, velocities and force buffers on the device.
130 * The coordinates buffer is (re)allocated, when required by PME, with a padding,
131 * the size of which is set by the constructor. The padding region clearing kernel
132 * is scheduled in the \p pmeStream_ (unlike the coordinates H2D) as only the PME
133 * task uses this padding area.
136 * The force buffer is cleared if its size increases, so that previously unused
137 * memory is cleared before forces are accumulated.
139 * \param[in] numAtomsLocal Number of atoms in local domain.
140 * \param[in] numAtomsAll Total number of atoms to handle.
142 void reinit(int numAtomsLocal, int numAtomsAll);
144 /*! \brief Returns the range of atoms to be copied based on the copy type (all, local or non-local).
146 * \todo There are at least three versions of the function with this functionality in the code:
147 * this one and two more in NBNXM. These should be unified in a shape of a general function
150 * \param[in] atomLocality If all, local or non-local ranges are needed.
152 * \returns Tuple, containing the index of the first atom in the range and the total number of atoms in the range.
154 std::tuple<int, int> getAtomRangesFromAtomLocality(AtomLocality atomLocality);
157 /*! \brief Get the positions buffer on the GPU.
159 * \returns GPU positions buffer.
161 DeviceBuffer<RVec> getCoordinates();
163 /*! \brief Copy positions to the GPU memory.
165 * \param[in] h_x Positions in the host memory.
166 * \param[in] atomLocality Locality of the particles to copy.
168 void copyCoordinatesToGpu(gmx::ArrayRef<const gmx::RVec> h_x, AtomLocality atomLocality);
170 /*! \brief Get the event synchronizer of the coordinates ready for the consumption on the device.
172 * Returns the event synchronizer which indicates that the coordinates are ready for the
173 * consumption on the device. Takes into account that the producer may be different.
175 * If the update is offloaded, and the current step is not a DD/search step, the returned
176 * synchronizer indicates the completion of GPU update-constraint kernels. Otherwise, on search
177 * steps and if update is not offloaded, the coordinates are provided by the H2D copy and the
178 * returned synchronizer indicates that the copy is complete.
180 * \param[in] atomLocality Locality of the particles to wait for.
181 * \param[in] simulationWork The simulation lifetime flags.
182 * \param[in] stepWork The step lifetime flags.
184 * \returns The event to synchronize the stream that consumes coordinates on device.
186 GpuEventSynchronizer* getCoordinatesReadyOnDeviceEvent(AtomLocality atomLocality,
187 const SimulationWorkload& simulationWork,
188 const StepWorkload& stepWork);
190 /*! \brief Blocking wait until coordinates are copied to the device.
192 * Synchronizes the stream in which the copy was executed.
194 * \param[in] atomLocality Locality of the particles to wait for.
196 void waitCoordinatesCopiedToDevice(AtomLocality atomLocality);
198 /*! \brief Getter for the event synchronizer for the update is done on th GPU
200 * \returns The event to synchronize the stream coordinates wre updated on device.
202 GpuEventSynchronizer* xUpdatedOnDevice();
204 /*! \brief Copy positions from the GPU memory.
206 * \param[in] h_x Positions buffer in the host memory.
207 * \param[in] atomLocality Locality of the particles to copy.
209 void copyCoordinatesFromGpu(gmx::ArrayRef<gmx::RVec> h_x, AtomLocality atomLocality);
211 /*! \brief Wait until coordinates are available on the host.
213 * \param[in] atomLocality Locality of the particles to wait for.
215 void waitCoordinatesReadyOnHost(AtomLocality atomLocality);
218 /*! \brief Get the velocities buffer on the GPU.
220 * \returns GPU velocities buffer.
222 DeviceBuffer<RVec> getVelocities();
224 /*! \brief Copy velocities to the GPU memory.
226 * \param[in] h_v Velocities in the host memory.
227 * \param[in] atomLocality Locality of the particles to copy.
229 void copyVelocitiesToGpu(gmx::ArrayRef<const gmx::RVec> h_v, AtomLocality atomLocality);
231 /*! \brief Get the event synchronizer on the H2D velocities copy.
233 * \param[in] atomLocality Locality of the particles to wait for.
235 * \returns The event to synchronize the stream that consumes velocities on device.
237 GpuEventSynchronizer* getVelocitiesReadyOnDeviceEvent(AtomLocality atomLocality);
239 /*! \brief Copy velocities from the GPU memory.
241 * \param[in] h_v Velocities buffer in the host memory.
242 * \param[in] atomLocality Locality of the particles to copy.
244 void copyVelocitiesFromGpu(gmx::ArrayRef<gmx::RVec> h_v, AtomLocality atomLocality);
246 /*! \brief Wait until velocities are available on the host.
248 * \param[in] atomLocality Locality of the particles to wait for.
250 void waitVelocitiesReadyOnHost(AtomLocality atomLocality);
253 /*! \brief Get the force buffer on the GPU.
255 * \returns GPU force buffer.
257 DeviceBuffer<RVec> getForces();
259 /*! \brief Copy forces to the GPU memory.
261 * \param[in] h_f Forces in the host memory.
262 * \param[in] atomLocality Locality of the particles to copy.
264 void copyForcesToGpu(gmx::ArrayRef<const gmx::RVec> h_f, AtomLocality atomLocality);
266 /*! \brief Get the event synchronizer for the forces ready on device.
268 * Returns either of the event synchronizers, depending on the offload scenario
269 * for the current simulation timestep:
270 * 1. The forces are copied to the device (when GPU buffer ops are off)
271 * 2. The forces are reduced on the device (GPU buffer ops are on)
273 * \todo Pass step workload instead of the useGpuFBufferOps boolean.
275 * \param[in] atomLocality Locality of the particles to wait for.
276 * \param[in] useGpuFBufferOps If the force buffer ops are offloaded to the GPU.
278 * \returns The event to synchronize the stream that consumes forces on device.
280 GpuEventSynchronizer* getForcesReadyOnDeviceEvent(AtomLocality atomLocality, bool useGpuFBufferOps);
282 /*! \brief Getter for the event synchronizer for the forces are reduced on the GPU.
284 * \returns The event to mark when forces are reduced on the GPU.
286 GpuEventSynchronizer* fReducedOnDevice();
288 /*! \brief Copy forces from the GPU memory.
290 * \param[in] h_f Forces buffer in the host memory.
291 * \param[in] atomLocality Locality of the particles to copy.
293 void copyForcesFromGpu(gmx::ArrayRef<gmx::RVec> h_f, AtomLocality atomLocality);
295 /*! \brief Wait until forces are available on the host.
297 * \param[in] atomLocality Locality of the particles to wait for.
299 void waitForcesReadyOnHost(AtomLocality atomLocality);
301 /*! \brief Getter for the update stream.
303 * \todo This is temporary here, until the management of this stream is taken over.
305 * \returns The device command stream to use in update-constraints.
307 const DeviceStream* getUpdateStream();
309 /*! \brief Getter for the number of local atoms.
311 * \returns The number of local atoms.
315 /*! \brief Getter for the total number of atoms.
317 * \returns The total number of atoms.
323 const DeviceStream* pmeStream_;
324 //! GPU NBNXM local stream.
325 const DeviceStream* localStream_;
326 //! GPU NBNXM non-local stream.
327 const DeviceStream* nonLocalStream_;
328 //! GPU Update-constreaints stream.
329 const DeviceStream* updateStream_;
331 // Streams to use for coordinates H2D and D2H copies (one event for each atom locality)
332 EnumerationArray<AtomLocality, const DeviceStream*> xCopyStreams_ = { { nullptr } };
333 // Streams to use for velocities H2D and D2H copies (one event for each atom locality)
334 EnumerationArray<AtomLocality, const DeviceStream*> vCopyStreams_ = { { nullptr } };
335 // Streams to use for forces H2D and D2H copies (one event for each atom locality)
336 EnumerationArray<AtomLocality, const DeviceStream*> fCopyStreams_ = { { nullptr } };
338 /*! \brief An array of events that indicate H2D copy is complete (one event for each atom locality)
340 * \todo Reconsider naming. It should be xCopiedToDevice or xH2DCopyComplete, etc.
342 EnumerationArray<AtomLocality, GpuEventSynchronizer> xReadyOnDevice_;
343 //! An event that the coordinates are ready after update-constraints execution
344 GpuEventSynchronizer xUpdatedOnDevice_;
345 //! An array of events that indicate D2H copy of coordinates is complete (one event for each atom locality)
346 EnumerationArray<AtomLocality, GpuEventSynchronizer> xReadyOnHost_;
348 //! An array of events that indicate H2D copy of velocities is complete (one event for each atom locality)
349 EnumerationArray<AtomLocality, GpuEventSynchronizer> vReadyOnDevice_;
350 //! An array of events that indicate D2H copy of velocities is complete (one event for each atom locality)
351 EnumerationArray<AtomLocality, GpuEventSynchronizer> vReadyOnHost_;
353 //! An array of events that indicate H2D copy of forces is complete (one event for each atom locality)
354 EnumerationArray<AtomLocality, GpuEventSynchronizer> fReadyOnDevice_;
355 //! An event that the forces were reduced on the GPU
356 GpuEventSynchronizer fReducedOnDevice_;
357 //! An array of events that indicate D2H copy of forces is complete (one event for each atom locality)
358 EnumerationArray<AtomLocality, GpuEventSynchronizer> fReadyOnHost_;
360 //! GPU context (for OpenCL builds)
361 const DeviceContext& deviceContext_;
362 //! Default GPU calls behavior
363 GpuApiCallBehavior transferKind_ = GpuApiCallBehavior::Async;
364 //! Required minimum divisor of the allocation size of the coordinates buffer
365 int allocationBlockSizeDivisor_ = 0;
367 //! Number of local atoms
368 int numAtomsLocal_ = -1;
369 //! Total number of atoms
370 int numAtomsAll_ = -1;
372 //! Device positions buffer
373 DeviceBuffer<RVec> d_x_;
374 //! Number of particles saved in the positions buffer
376 //! Allocation size for the positions buffer
377 int d_xCapacity_ = -1;
379 //! Device velocities buffer
380 DeviceBuffer<RVec> d_v_;
381 //! Number of particles saved in the velocities buffer
383 //! Allocation size for the velocities buffer
384 int d_vCapacity_ = -1;
386 //! Device force buffer
387 DeviceBuffer<RVec> d_f_;
388 //! Number of particles saved in the force buffer
390 //! Allocation size for the force buffer
391 int d_fCapacity_ = -1;
393 //! \brief Pointer to wallcycle structure.
394 gmx_wallcycle* wcycle_;
396 /*! \brief Performs the copy of data from host to device buffer.
398 * \todo Template on locality.
400 * \param[out] d_data Device-side buffer.
401 * \param[in] h_data Host-side buffer.
402 * \param[in] dataSize Device-side data allocation size.
403 * \param[in] atomLocality If all, local or non-local ranges should be copied.
404 * \param[in] deviceStream GPU stream to execute copy in.
406 void copyToDevice(DeviceBuffer<RVec> d_data,
407 gmx::ArrayRef<const gmx::RVec> h_data,
409 AtomLocality atomLocality,
410 const DeviceStream& deviceStream);
412 /*! \brief Performs the copy of data from device to host buffer.
414 * \param[out] h_data Host-side buffer.
415 * \param[in] d_data Device-side buffer.
416 * \param[in] dataSize Device-side data allocation size.
417 * \param[in] atomLocality If all, local or non-local ranges should be copied.
418 * \param[in] deviceStream GPU stream to execute copy in.
420 void copyFromDevice(gmx::ArrayRef<gmx::RVec> h_data,
421 DeviceBuffer<RVec> d_data,
423 AtomLocality atomLocality,
424 const DeviceStream& deviceStream);
429 #endif // GMX_MDTYPES_STATE_PROPAGATOR_DATA_GPU_IMPL_H