2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 1991-2000, University of Groningen, The Netherlands.
5 * Copyright (c) 2001-2012, The GROMACS development team.
6 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
7 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
8 * and including many others, as listed in the AUTHORS file in the
9 * top-level source directory and at http://www.gromacs.org.
11 * GROMACS is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public License
13 * as published by the Free Software Foundation; either version 2.1
14 * of the License, or (at your option) any later version.
16 * GROMACS is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with GROMACS; if not, see
23 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
24 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
26 * If you want to redistribute modifications to GROMACS, please
27 * consider that scientific software is very special. Version
28 * control is crucial - bugs must be traceable. We will be happy to
29 * consider code for inclusion in the official distribution, but
30 * derived work must not be called official GROMACS. Details are found
31 * in the README & COPYING files - if they are missing, get the
32 * official version at http://www.gromacs.org.
34 * To help us fund GROMACS development, we humbly ask that you cite
35 * the research papers on the package. Check out http://www.gromacs.org.
40 * Data types used internally in the nbnxn_cuda module.
42 * \author Szilárd Páll <pall.szilard@gmail.com>
43 * \ingroup module_mdlib
46 #ifndef NBNXN_CUDA_TYPES_H
47 #define NBNXN_CUDA_TYPES_H
49 #include "gromacs/legacyheaders/types/interaction_const.h"
50 #include "gromacs/mdlib/nbnxn_pairlist.h"
51 #include "gromacs/legacyheaders/types/nbnxn_cuda_types_ext.h"
52 #include "../../gmxlib/cuda_tools/cudautils.cuh"
54 /* CUDA versions from 5.0 above support texture objects. */
55 #if CUDA_VERSION >= 5000
56 #define TEXOBJ_SUPPORTED
57 #else /* CUDA_VERSION */
58 /** This typedef allows us to define only one version of struct cu_nbparam */
59 typedef int cudaTextureObject_t;
60 #endif /* CUDA_VERSION */
66 /*! \brief Electrostatic CUDA kernel flavors.
68 * Types of electrostatics implementations available in the CUDA non-bonded
69 * force kernels. These represent both the electrostatics types implemented
70 * by the kernels (cut-off, RF, and Ewald - a subset of what's defined in
71 * enums.h) as well as encode implementation details analytical/tabulated
72 * and single or twin cut-off (for Ewald kernels).
73 * Note that the cut-off and RF kernels have only analytical flavor and unlike
74 * in the CPU kernels, the tabulated kernels are ATM Ewald-only.
76 * The row-order of pointers to different electrostatic kernels defined in
77 * nbnxn_cuda.cu by the nb_*_kfunc_ptr function pointer table
78 * should match the order of enumerated types below.
81 eelCuCUT, eelCuRF, eelCuEWALD_TAB, eelCuEWALD_TAB_TWIN, eelCuEWALD_ANA, eelCuEWALD_ANA_TWIN, eelCuNR
84 /*! \brief VdW CUDA kernel flavors.
86 * The enumerates values correspond to the LJ implementations in the CUDA non-bonded
89 * The column-order of pointers to different electrostatic kernels defined in
90 * nbnxn_cuda.cu by the nb_*_kfunc_ptr function pointer table
91 * should match the order of enumerated types below.
94 evdwCuCUT, evdwCuFSWITCH, evdwCuPSWITCH, evdwCuEWALDGEOM, evdwCuEWALDLB, evdwCuNR
97 /* All structs prefixed with "cu_" hold data used in GPU calculations and
98 * are passed to the kernels, except cu_timers_t. */
100 typedef struct cu_plist cu_plist_t;
101 typedef struct cu_atomdata cu_atomdata_t;
102 typedef struct cu_nbparam cu_nbparam_t;
103 typedef struct cu_timers cu_timers_t;
104 typedef struct nb_staging nb_staging_t;
109 * \brief Staging area for temporary data downloaded from the GPU.
111 * The energies/shift forces get downloaded here first, before getting added
112 * to the CPU-side aggregate values.
116 float *e_lj; /**< LJ energy */
117 float *e_el; /**< electrostatic energy */
118 float3 *fshift; /**< shift forces */
122 * \brief Nonbonded atom data - both inputs and outputs.
126 int natoms; /**< number of atoms */
127 int natoms_local; /**< number of local atoms */
128 int nalloc; /**< allocation size for the atom data (xq, f) */
130 float4 *xq; /**< atom coordinates + charges, size natoms */
131 float3 *f; /**< force output array, size natoms */
133 float *e_lj; /**< LJ energy output, size 1 */
134 float *e_el; /**< Electrostatics energy input, size 1 */
136 float3 *fshift; /**< shift forces */
138 int ntypes; /**< number of atom types */
139 int *atom_types; /**< atom type indices, size natoms */
141 float3 *shift_vec; /**< shifts */
142 bool bShiftVecUploaded; /**< true if the shift vector has been uploaded */
146 * \brief Parameters required for the CUDA nonbonded calculations.
151 int eeltype; /**< type of electrostatics, takes values from #eelCu */
152 int vdwtype; /**< type of VdW impl., takes values from #evdwCu */
154 float epsfac; /**< charge multiplication factor */
155 float c_rf; /**< Reaction-field/plain cutoff electrostatics const. */
156 float two_k_rf; /**< Reaction-field electrostatics constant */
157 float ewald_beta; /**< Ewald/PME parameter */
158 float sh_ewald; /**< Ewald/PME correction term substracted from the direct-space potential */
159 float sh_lj_ewald; /**< LJ-Ewald/PME correction term added to the correction potential */
160 float ewaldcoeff_lj; /**< LJ-Ewald/PME coefficient */
162 float rcoulomb_sq; /**< Coulomb cut-off squared */
164 float rvdw_sq; /**< VdW cut-off squared */
165 float rvdw_switch; /**< VdW switched cut-off */
166 float rlist_sq; /**< pair-list cut-off squared */
168 shift_consts_t dispersion_shift; /**< VdW shift dispersion constants */
169 shift_consts_t repulsion_shift; /**< VdW shift repulsion constants */
170 switch_consts_t vdw_switch; /**< VdW switch constants */
172 /* LJ non-bonded parameters - accessed through texture memory */
173 float *nbfp; /**< nonbonded parameter table with C6/C12 pairs per atom type-pair, 2*ntype^2 elements */
174 cudaTextureObject_t nbfp_texobj; /**< texture object bound to nbfp */
175 float *nbfp_comb; /**< nonbonded parameter table per atom type, 2*ntype elements */
176 cudaTextureObject_t nbfp_comb_texobj; /**< texture object bound to nbfp_texobj */
178 /* Ewald Coulomb force table data - accessed through texture memory */
179 int coulomb_tab_size; /**< table size (s.t. it fits in texture cache) */
180 float coulomb_tab_scale; /**< table scale/spacing */
181 float *coulomb_tab; /**< pointer to the table in the device memory */
182 cudaTextureObject_t coulomb_tab_texobj; /**< texture object bound to coulomb_tab */
186 * \brief Pair list data.
190 int na_c; /**< number of atoms per cluster */
192 int nsci; /**< size of sci, # of i clusters in the list */
193 int sci_nalloc; /**< allocation size of sci */
194 nbnxn_sci_t *sci; /**< list of i-cluster ("super-clusters") */
196 int ncj4; /**< total # of 4*j clusters */
197 int cj4_nalloc; /**< allocation size of cj4 */
198 nbnxn_cj4_t *cj4; /**< 4*j cluster list, contains j cluster number
199 and index into the i cluster list */
200 nbnxn_excl_t *excl; /**< atom interaction bits */
201 int nexcl; /**< count for excl */
202 int excl_nalloc; /**< allocation size of excl */
204 bool bDoPrune; /**< true if pair-list pruning needs to be
205 done during the current step */
209 * \brief CUDA events used for timing GPU kernels and H2D/D2H transfers.
211 * The two-sized arrays hold the local and non-local values and should always
212 * be indexed with eintLocal/eintNonlocal.
216 cudaEvent_t start_atdat; /**< start event for atom data transfer (every PS step) */
217 cudaEvent_t stop_atdat; /**< stop event for atom data transfer (every PS step) */
218 cudaEvent_t start_nb_h2d[2]; /**< start events for x/q H2D transfers (l/nl, every step) */
219 cudaEvent_t stop_nb_h2d[2]; /**< stop events for x/q H2D transfers (l/nl, every step) */
220 cudaEvent_t start_nb_d2h[2]; /**< start events for f D2H transfer (l/nl, every step) */
221 cudaEvent_t stop_nb_d2h[2]; /**< stop events for f D2H transfer (l/nl, every step) */
222 cudaEvent_t start_pl_h2d[2]; /**< start events for pair-list H2D transfers (l/nl, every PS step) */
223 cudaEvent_t stop_pl_h2d[2]; /**< start events for pair-list H2D transfers (l/nl, every PS step) */
224 cudaEvent_t start_nb_k[2]; /**< start event for non-bonded kernels (l/nl, every step) */
225 cudaEvent_t stop_nb_k[2]; /**< stop event non-bonded kernels (l/nl, every step) */
229 * \brief Main data structure for CUDA nonbonded force calculations.
233 cuda_dev_info_t *dev_info; /**< CUDA device information */
234 bool bUseTwoStreams; /**< true if doing both local/non-local NB work on GPU */
235 bool bUseStreamSync; /**< true if the standard cudaStreamSynchronize is used
236 and not memory polling-based waiting */
237 cu_atomdata_t *atdat; /**< atom data */
238 cu_nbparam_t *nbparam; /**< parameters required for the non-bonded calc. */
239 cu_plist_t *plist[2]; /**< pair-list data structures (local and non-local) */
240 nb_staging_t nbst; /**< staging area where fshift/energies get downloaded */
242 cudaStream_t stream[2]; /**< local and non-local GPU streams */
244 /** events used for synchronization */
245 cudaEvent_t nonlocal_done; /**< event triggered when the non-local non-bonded kernel
246 is done (and the local transfer can proceed) */
247 cudaEvent_t misc_ops_done; /**< event triggered when the operations that precede the
248 main force calculations are done (e.g. buffer 0-ing) */
250 /* NOTE: With current CUDA versions (<=5.0) timing doesn't work with multiple
251 * concurrent streams, so we won't time if both l/nl work is done on GPUs.
252 * Timer init/uninit is still done even with timing off so only the condition
253 * setting bDoTime needs to be change if this CUDA "feature" gets fixed. */
254 bool bDoTime; /**< True if event-based timing is enabled. */
255 cu_timers_t *timers; /**< CUDA event-based timers. */
256 wallclock_gpu_t *timings; /**< Timing data. */
263 #endif /* NBNXN_CUDA_TYPES_H */