2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2018,2019, 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.
36 * \brief Declares GPU implementation class for CUDA bonded
39 * This header file is needed to include from both the device-side
40 * kernels file, and the host-side management code.
42 * \author Berk Hess <hess@kth.se>
43 * \author Szilárd Páll <pall.szilard@gmail.com>
44 * \author Mark Abraham <mark.j.abraham@gmail.com>
46 * \ingroup module_listed_forces
48 #ifndef GMX_LISTED_FORCES_GPUBONDED_IMPL_H
49 #define GMX_LISTED_FORCES_GPUBONDED_IMPL_H
51 #include "gromacs/gpu_utils/gpu_vec.cuh"
52 #include "gromacs/gpu_utils/gputraits.cuh"
53 #include "gromacs/gpu_utils/hostallocator.h"
54 #include "gromacs/listed_forces/gpubonded.h"
55 #include "gromacs/pbcutil/pbc_aiuc.h"
56 #include "gromacs/topology/idef.h"
58 struct gmx_ffparams_t;
64 /*! \internal \brief Version of InteractionList that supports pinning */
65 struct HostInteractionList
67 /*! \brief Returns the total number of elements in iatoms */
73 //! List of interactions, see \c HostInteractionLists
74 HostVector<int> iatoms = {{}, gmx::HostAllocationPolicy(gmx::PinningPolicy::PinnedIfSupported)};
77 /* \brief Bonded parameters and GPU pointers
79 * This is used to accumulate all the parameters and pointers so they can be passed
80 * to the GPU as a single structure.
83 struct BondedCudaKernelParameters
85 //! Periodic boundary data
89 //! The bonded types on GPU
90 int ftypesOnGpu[nFtypesOnGpu];
91 //! The number of interaction atom (iatom) elements for every function type
92 int nrFTypeIAtoms[nFtypesOnGpu];
93 //! The number of bonds for every function type
94 int nrFTypeBonds[nFtypesOnGpu];
95 //! The start index in the range of each interaction type
96 int ftypeRangeStart[nFtypesOnGpu];
97 //! The end index in the range of each interaction type
98 int ftypeRangeEnd[nFtypesOnGpu];
100 //! Force parameters (on GPU)
101 t_iparams *forceparamsDevice;
102 //! Coordinates before the timestep (on GPU)
103 const float4 *xqDevice;
104 //! Forces on atoms (on GPU)
106 //! Force shifts on atoms (on GPU)
108 //! Total Energy (on GPU)
110 //! Interaction list atoms (on GPU)
111 t_iatom *iatoms[nFtypesOnGpu];
113 BondedCudaKernelParameters()
115 matrix boxDummy = { {0, 0, 0}, {0, 0, 0}, {0, 0, 0} };
117 setPbcAiuc(0, boxDummy, &pbcAiuc);
120 forceparamsDevice = nullptr;
122 forceDevice = nullptr;
123 fshiftDevice = nullptr;
124 vtotDevice = nullptr;
128 /*! \internal \brief Implements GPU bondeds */
129 class GpuBonded::Impl
133 Impl(const gmx_ffparams_t &ffparams,
135 /*! \brief Destructor, non-default needed for freeing
136 * device-side buffers */
138 /*! \brief Update lists of interactions from idef suitable for the GPU,
139 * using the data structures prepared for PP work.
141 * Intended to be called after each neighbour search
142 * stage. Copies the bonded interactions assigned to the GPU
143 * to device data structures, and updates device buffers that
144 * may have been updated after search. */
145 void updateInteractionListsAndDeviceBuffers(ArrayRef<const int> nbnxnAtomOrder,
151 /*! \brief Launches bonded kernel on a GPU */
152 template <bool calcVir, bool calcEner>
154 launchKernel(const t_forcerec *fr,
156 /*! \brief Returns whether there are bonded interactions
157 * assigned to the GPU */
158 bool haveInteractions() const;
159 /*! \brief Launches the transfer of computed bonded energies. */
160 void launchEnergyTransfer();
161 /*! \brief Waits on the energy transfer, and accumulates bonded energies to \c enerd. */
162 void accumulateEnergyTerms(gmx_enerdata_t *enerd);
163 /*! \brief Clears the device side energy buffer */
164 void clearEnergies();
166 /*! \brief The interaction lists
168 * \todo This is potentially several pinned allocations, which
169 * could contribute to exhausting such pages. */
170 std::array<HostInteractionList, F_NRE> iLists;
171 //! Tells whether there are any interaction in iLists.
172 bool haveInteractions_;
173 //! Interaction lists on the device.
174 t_ilist iListsDevice[F_NRE];
176 //! Bonded parameters for device-side use.
177 t_iparams *forceparamsDevice = nullptr;
178 //! Position-charge vector on the device.
179 const float4 *xqDevice = nullptr;
180 //! Force vector on the device.
181 fvec *forceDevice = nullptr;
182 //! Shift force vector on the device.
183 fvec *fshiftDevice = nullptr;
184 //! \brief Host-side virial buffer
185 HostVector <float> vtot = {{}, gmx::HostAllocationPolicy(gmx::PinningPolicy::PinnedIfSupported)};
186 //! \brief Device-side total virial
187 float *vtotDevice = nullptr;
189 //! \brief Bonded GPU stream, not owned by this module
190 CommandStream stream;
192 //! Parameters and pointers, passed to the CUDA kernel
193 BondedCudaKernelParameters kernelParams_;