2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2016 by the GROMACS development team.
5 * Copyright (c) 2017,2018,2019,2020,2021, by the GROMACS development team, led by
6 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
7 * and including many others, as listed in the AUTHORS file in the
8 * top-level source directory and at http://www.gromacs.org.
10 * GROMACS is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public License
12 * as published by the Free Software Foundation; either version 2.1
13 * of the License, or (at your option) any later version.
15 * GROMACS is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with GROMACS; if not, see
22 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
23 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 * If you want to redistribute modifications to GROMACS, please
26 * consider that scientific software is very special. Version
27 * control is crucial - bugs must be traceable. We will be happy to
28 * consider code for inclusion in the official distribution, but
29 * derived work must not be called official GROMACS. Details are found
30 * in the README & COPYING files - if they are missing, get the
31 * official version at http://www.gromacs.org.
33 * To help us fund GROMACS development, we humbly ask that you cite
34 * the research papers on the package. Check out http://www.gromacs.org.
39 * Data types used internally in the nbnxm_ocl module.
41 * \author Anca Hamuraru <anca@streamcomputing.eu>
42 * \author Szilárd Páll <pszilard@kth.se>
43 * \ingroup module_nbnxm
46 #ifndef GMX_NBNXM_NBNXM_OPENCL_TYPES_H
47 #define GMX_NBNXM_NBNXM_OPENCL_TYPES_H
49 #include "gromacs/gpu_utils/devicebuffer.h"
50 #include "gromacs/gpu_utils/gmxopencl.h"
51 #include "gromacs/gpu_utils/gpueventsynchronizer_ocl.h"
52 #include "gromacs/gpu_utils/gputraits_ocl.h"
53 #include "gromacs/gpu_utils/oclutils.h"
54 #include "gromacs/mdtypes/interaction_const.h"
55 #include "gromacs/nbnxm/gpu_types_common.h"
56 #include "gromacs/nbnxm/nbnxm.h"
57 #include "gromacs/nbnxm/nbnxm_gpu.h"
58 #include "gromacs/nbnxm/pairlist.h"
59 #include "gromacs/utility/enumerationhelpers.h"
60 #include "gromacs/utility/fatalerror.h"
61 #include "gromacs/utility/real.h"
63 struct gmx_wallclock_gpu_nbnxn_t;
65 /*! \brief Pruning kernel flavors.
67 * The values correspond to the first call of the pruning post-list generation
68 * and the rolling pruning, respectively.
78 * \brief Data structure shared between the OpenCL device code and OpenCL host code
80 * Must not contain OpenCL objects (buffers)
81 * TODO: review, improve */
82 typedef struct cl_nbparam_params
85 //! type of electrostatics
86 enum Nbnxm::ElecType elecType;
88 enum Nbnxm::VdwType vdwType;
90 //! charge multiplication factor
92 //! Reaction-field/plain cutoff electrostatics const.
94 //! Reaction-field electrostatics constant
96 //! Ewald/PME parameter
98 //! Ewald/PME correction term subtracted from the direct-space potential
100 //! LJ-Ewald/PME correction term added to the correction potential
102 //! LJ-Ewald/PME coefficient
105 //! Coulomb cut-off squared
108 //! VdW cut-off squared
110 //! VdW switched cut-off
112 //! Full, outer pair-list cut-off squared
114 //! Inner, dynamic pruned pair-list cut-off squared
117 //! VdW shift dispersion constants
118 shift_consts_t dispersion_shift;
119 //! VdW shift repulsion constants
120 shift_consts_t repulsion_shift;
121 //! VdW switch constants
122 switch_consts_t vdw_switch;
124 /* Ewald Coulomb force table data - accessed through texture memory */
125 //! table scale/spacing
126 float coulomb_tab_scale;
127 } cl_nbparam_params_t;
131 * \brief Main data structure for OpenCL nonbonded force calculations.
135 /* \brief OpenCL device context
137 * \todo Make it constant reference, once NbnxmGpu is a proper class.
139 const DeviceContext* deviceContext_;
140 //! OpenCL runtime data (context, kernels)
141 struct gmx_device_runtime_data_t* dev_rundata = nullptr;
143 /**< Pointers to non-bonded kernel functions
144 * organized similar with nb_kfunc_xxx arrays in nbnxn_ocl.cpp */
146 cl_kernel kernel_noener_noprune_ptr[Nbnxm::c_numElecTypes][Nbnxm::c_numVdwTypes] = { { nullptr } };
147 cl_kernel kernel_ener_noprune_ptr[Nbnxm::c_numElecTypes][Nbnxm::c_numVdwTypes] = { { nullptr } };
148 cl_kernel kernel_noener_prune_ptr[Nbnxm::c_numElecTypes][Nbnxm::c_numVdwTypes] = { { nullptr } };
149 cl_kernel kernel_ener_prune_ptr[Nbnxm::c_numElecTypes][Nbnxm::c_numVdwTypes] = { { nullptr } };
151 //! prune kernels, ePruneKind defined the kernel kinds
152 cl_kernel kernel_pruneonly[ePruneNR] = { nullptr };
154 //! true if prefetching fg i-atom LJ parameters should be used in the kernels
155 bool bPrefetchLjParam = false;
157 /**< auxiliary kernels implementing memset-like functions */
159 cl_kernel kernel_memset_f = nullptr;
160 cl_kernel kernel_memset_f2 = nullptr;
161 cl_kernel kernel_memset_f3 = nullptr;
164 //! true if doing both local/non-local NB work on GPU
165 bool bUseTwoStreams = false;
166 //! true indicates that the nonlocal_done event was marked
167 bool bNonLocalStreamDoneMarked = false;
170 NBAtomData* atdat = nullptr;
171 //! parameters required for the non-bonded calc.
172 NBParamGpu* nbparam = nullptr;
173 //! pair-list data structures (local and non-local)
174 gmx::EnumerationArray<Nbnxm::InteractionLocality, Nbnxm::gpu_plist*> plist = { nullptr };
175 //! staging area where fshift/energies get downloaded
178 // Data for GPU-side coordinate conversion between integrator and NBNXM
179 /*! \brief array of atom indices */
180 DeviceBuffer<int> atomIndices;
181 /*! \brief size of atom indices */
182 int atomIndicesSize = 0;
183 /*! \brief size of atom indices allocated in device buffer */
184 int atomIndicesSize_alloc = 0;
185 /*! \brief x buf ops num of atoms */
186 DeviceBuffer<int> cxy_na;
187 /*! \brief number of elements in cxy_na */
189 /*! \brief number of elements allocated allocated in device buffer */
190 int ncxy_na_alloc = 0;
191 /*! \brief x buf ops cell index mapping */
192 DeviceBuffer<int> cxy_ind;
193 /*! \brief number of elements in cxy_ind */
195 /*! \brief number of elements allocated allocated in device buffer */
196 int ncxy_ind_alloc = 0;
198 //! local and non-local GPU queues
199 gmx::EnumerationArray<Nbnxm::InteractionLocality, const DeviceStream*> deviceStreams;
201 /*! \brief Events used for synchronization */
203 /*! \brief Event triggered when the non-local non-bonded
204 * kernel is done (and the local transfer can proceed) */
205 GpuEventSynchronizer nonlocal_done;
206 /*! \brief Event triggered when the tasks issued in the local
207 * stream that need to precede the non-local force or buffer
208 * operation calculations are done (e.g. f buffer 0-ing, local
209 * x/q H2D, buffer op initialization in local stream that is
210 * required also by nonlocal stream ) */
211 GpuEventSynchronizer misc_ops_and_local_H2D_done;
214 //! True if there has been local/nonlocal GPU work, either bonded or nonbonded, scheduled
215 // to be executed in the current domain. As long as bonded work is not split up into
216 // local/nonlocal, if there is bonded GPU work, both flags will be true.
217 gmx::EnumerationArray<Nbnxm::InteractionLocality, bool> haveWork;
220 //! True if event-based timing is enabled.
221 bool bDoTime = false;
222 //! OpenCL event-based timers.
223 Nbnxm::GpuTimers* timers = nullptr;
224 //! Timing data. TODO: deprecate this and query timers for accumulated data instead
225 gmx_wallclock_gpu_nbnxn_t* timings = nullptr;
228 #endif /* NBNXN_OPENCL_TYPES_H */