2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2017,2018,2019,2020,2021, 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.
36 * \brief Implements common internal types for different NBNXN GPU implementations
38 * \author Szilárd Páll <pall.szilard@gmail.com>
39 * \ingroup module_nbnxm
42 #ifndef GMX_MDLIB_NBNXN_GPU_COMMON_TYPES_H
43 #define GMX_MDLIB_NBNXN_GPU_COMMON_TYPES_H
47 #include "gromacs/mdtypes/interaction_const.h"
48 #include "gromacs/mdtypes/locality.h"
49 #include "gromacs/utility/enumerationhelpers.h"
55 # include "gromacs/gpu_utils/gpuregiontimer_ocl.h"
59 # include "gromacs/gpu_utils/gpuregiontimer.cuh"
63 # include "gromacs/gpu_utils/gpuregiontimer_sycl.h"
66 /*! \brief Macro definining default for the prune kernel's j4 processing concurrency.
68 * The GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY macro allows compile-time override with the default value of 4.
70 #ifndef GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY
71 # define GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY 4
73 //! Default for the prune kernel's j4 processing concurrency.
74 static constexpr int c_pruneKernelJ4Concurrency = GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY;
77 * \brief Parameters required for the GPU nonbonded calculations.
82 //! type of electrostatics
83 enum Nbnxm::ElecType elecType;
85 enum Nbnxm::VdwType vdwType;
87 //! charge multiplication factor
89 //! Reaction-field/plain cutoff electrostatics const.
91 //! Reaction-field electrostatics constant
93 //! Ewald/PME parameter
95 //! Ewald/PME correction term subtracted from the direct-space potential
97 //! LJ-Ewald/PME correction term added to the correction potential
99 //! LJ-Ewald/PME coefficient
102 //! Coulomb cut-off squared
105 //! VdW cut-off squared
107 //! VdW switched cut-off
109 //! Full, outer pair-list cut-off squared
111 //! Inner, dynamic pruned pair-list cut-off squared
113 //! True if we use dynamic pair-list pruning
114 bool useDynamicPruning;
116 //! VdW shift dispersion constants
117 shift_consts_t dispersion_shift;
118 //! VdW shift repulsion constants
119 shift_consts_t repulsion_shift;
120 //! VdW switch constants
121 switch_consts_t vdw_switch;
123 /* LJ non-bonded parameters - accessed through texture memory */
124 //! nonbonded parameter table with C6/C12 pairs per atom type-pair, 2*ntype^2 elements
125 DeviceBuffer<float> nbfp;
126 //! texture object bound to nbfp
127 DeviceTexture nbfp_texobj;
128 //! nonbonded parameter table per atom type, 2*ntype elements
129 DeviceBuffer<float> nbfp_comb;
130 //! texture object bound to nbfp_comb
131 DeviceTexture nbfp_comb_texobj;
133 /* Ewald Coulomb force table data - accessed through texture memory */
134 //! table scale/spacing
135 float coulomb_tab_scale;
136 //! pointer to the table in the device memory
137 DeviceBuffer<float> coulomb_tab;
138 //! texture object bound to coulomb_tab
139 DeviceTexture coulomb_tab_texobj;
145 using gmx::AtomLocality;
146 using gmx::InteractionLocality;
149 * \brief GPU region timers used for timing GPU kernels and H2D/D2H transfers.
151 * The two-sized arrays hold the local and non-local values and should always
152 * be indexed with eintLocal/eintNonlocal.
157 * \brief Timers for local or non-local coordinate/force transfers
161 //! timer for x/q H2D transfers (l/nl, every step)
162 GpuRegionTimer nb_h2d;
163 //! timer for f D2H transfer (l/nl, every step)
164 GpuRegionTimer nb_d2h;
168 * \brief Timers for local or non-local interaction related operations
172 //! timer for pair-list H2D transfers (l/nl, every PS step)
173 GpuRegionTimer pl_h2d;
174 //! true when a pair-list transfer has been done at this step
175 bool didPairlistH2D = false;
176 //! timer for non-bonded kernels (l/nl, every step)
178 //! timer for the 1st pass list pruning kernel (l/nl, every PS step)
179 GpuRegionTimer prune_k;
180 //! true when we timed pruning and the timings need to be accounted for
181 bool didPrune = false;
182 //! timer for rolling pruning kernels (l/nl, frequency depends on chunk size)
183 GpuRegionTimer rollingPrune_k;
184 //! true when we timed rolling pruning (at the previous step) and the timings need to be accounted for
185 bool didRollingPrune = false;
188 //! timer for atom data transfer (every PS step)
189 GpuRegionTimer atdat;
190 //! timers for coordinate/force transfers (every step)
191 gmx::EnumerationArray<AtomLocality, XFTransfers> xf;
192 //! timers for interaction related transfers
193 gmx::EnumerationArray<InteractionLocality, Nbnxm::gpu_timers_t::Interaction> interaction;
197 * \brief GPU pair list structure */
200 //! number of atoms per cluster
203 //! size of sci, # of i clusters in the list
205 //! allocation size of sci
207 //! list of i-cluster ("super-clusters")
208 DeviceBuffer<nbnxn_sci_t> sci;
210 //! total # of 4*j clusters
212 //! allocation size of cj4
214 //! 4*j cluster list, contains j cluster number and index into the i cluster list
215 DeviceBuffer<nbnxn_cj4_t> cj4;
216 //! # of 4*j clusters * # of warps
218 //! allocation size of imask
220 //! imask for 2 warps for each 4*j cluster group
221 DeviceBuffer<unsigned int> imask;
222 //! atom interaction bits
223 DeviceBuffer<nbnxn_excl_t> excl;
226 //! allocation size of excl
229 /* parameter+variables for normal and rolling pruning */
230 //! true after search, indicates that initial pruning with outer pruning is needed
232 //! the number of parts/steps over which one cycle of rolling pruning takes places
233 int rollingPruningNumParts;
234 //! the next part to which the rolling pruning needs to be applied
235 int rollingPruningPart;