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) 2012,2013,2014,2015,2016,2017, 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/gpu_utils/cuda_arch_utils.cuh"
50 #include "gromacs/gpu_utils/cudautils.cuh"
51 #include "gromacs/mdlib/nbnxn_consts.h"
52 #include "gromacs/mdlib/nbnxn_pairlist.h"
53 #include "gromacs/mdtypes/interaction_const.h"
54 #include "gromacs/timing/gpu_timing.h"
57 /*! \brief Macro definining default for the prune kernel's j4 processing concurrency.
59 * The GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY macro allows compile-time override.
61 #ifndef GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY
62 #define GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY 4
64 /*! \brief Default for the prune kernel's j4 processing concurrency.
66 * Initialized using the #GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY macro which allows compile-time override.
68 const int c_cudaPruneKernelJ4Concurrency = GMX_NBNXN_PRUNE_KERNEL_J4_CONCURRENCY;
70 /* TODO: consider moving this to kernel_utils */
71 /* Convenience defines */
72 /*! \brief number of clusters per supercluster. */
73 static const int c_numClPerSupercl = c_nbnxnGpuNumClusterPerSupercluster;
74 /*! \brief cluster size = number of atoms per cluster. */
75 static const int c_clSize = c_nbnxnGpuClusterSize;
77 /*! \brief True if the use of texture fetch in the CUDA kernels is disabled. */
78 static const bool c_disableCudaTextures = DISABLE_CUDA_TEXTURES;
85 /*! \brief Electrostatic CUDA kernel flavors.
87 * Types of electrostatics implementations available in the CUDA non-bonded
88 * force kernels. These represent both the electrostatics types implemented
89 * by the kernels (cut-off, RF, and Ewald - a subset of what's defined in
90 * enums.h) as well as encode implementation details analytical/tabulated
91 * and single or twin cut-off (for Ewald kernels).
92 * Note that the cut-off and RF kernels have only analytical flavor and unlike
93 * in the CPU kernels, the tabulated kernels are ATM Ewald-only.
95 * The row-order of pointers to different electrostatic kernels defined in
96 * nbnxn_cuda.cu by the nb_*_kfunc_ptr function pointer table
97 * should match the order of enumerated types below.
100 eelCuCUT, eelCuRF, eelCuEWALD_TAB, eelCuEWALD_TAB_TWIN, eelCuEWALD_ANA, eelCuEWALD_ANA_TWIN, eelCuNR
103 /*! \brief VdW CUDA kernel flavors.
105 * The enumerates values correspond to the LJ implementations in the CUDA non-bonded
108 * The column-order of pointers to different electrostatic kernels defined in
109 * nbnxn_cuda.cu by the nb_*_kfunc_ptr function pointer table
110 * should match the order of enumerated types below.
113 evdwCuCUT, evdwCuCUTCOMBGEOM, evdwCuCUTCOMBLB, evdwCuFSWITCH, evdwCuPSWITCH, evdwCuEWALDGEOM, evdwCuEWALDLB, evdwCuNR
116 /* All structs prefixed with "cu_" hold data used in GPU calculations and
117 * are passed to the kernels, except cu_timers_t. */
119 typedef struct cu_plist cu_plist_t;
120 typedef struct cu_atomdata cu_atomdata_t;
121 typedef struct cu_nbparam cu_nbparam_t;
122 typedef struct cu_timers cu_timers_t;
123 typedef struct nb_staging nb_staging_t;
128 * \brief Staging area for temporary data downloaded from the GPU.
130 * The energies/shift forces get downloaded here first, before getting added
131 * to the CPU-side aggregate values.
135 float *e_lj; /**< LJ energy */
136 float *e_el; /**< electrostatic energy */
137 float3 *fshift; /**< shift forces */
141 * \brief Nonbonded atom data - both inputs and outputs.
145 int natoms; /**< number of atoms */
146 int natoms_local; /**< number of local atoms */
147 int nalloc; /**< allocation size for the atom data (xq, f) */
149 float4 *xq; /**< atom coordinates + charges, size natoms */
150 float3 *f; /**< force output array, size natoms */
152 float *e_lj; /**< LJ energy output, size 1 */
153 float *e_el; /**< Electrostatics energy input, size 1 */
155 float3 *fshift; /**< shift forces */
157 int ntypes; /**< number of atom types */
158 int *atom_types; /**< atom type indices, size natoms */
159 float2 *lj_comb; /**< sqrt(c6),sqrt(c12) size natoms */
161 float3 *shift_vec; /**< shifts */
162 bool bShiftVecUploaded; /**< true if the shift vector has been uploaded */
166 * \brief Parameters required for the CUDA nonbonded calculations.
171 int eeltype; /**< type of electrostatics, takes values from #eelCu */
172 int vdwtype; /**< type of VdW impl., takes values from #evdwCu */
174 float epsfac; /**< charge multiplication factor */
175 float c_rf; /**< Reaction-field/plain cutoff electrostatics const. */
176 float two_k_rf; /**< Reaction-field electrostatics constant */
177 float ewald_beta; /**< Ewald/PME parameter */
178 float sh_ewald; /**< Ewald/PME correction term substracted from the direct-space potential */
179 float sh_lj_ewald; /**< LJ-Ewald/PME correction term added to the correction potential */
180 float ewaldcoeff_lj; /**< LJ-Ewald/PME coefficient */
182 float rcoulomb_sq; /**< Coulomb cut-off squared */
184 float rvdw_sq; /**< VdW cut-off squared */
185 float rvdw_switch; /**< VdW switched cut-off */
186 float rlistOuter_sq; /**< Full, outer pair-list cut-off squared */
187 float rlistInner_sq; /**< Inner, dynamic pruned pair-list cut-off squared */
188 bool useDynamicPruning; /**< True if we use dynamic pair-list pruning */
190 shift_consts_t dispersion_shift; /**< VdW shift dispersion constants */
191 shift_consts_t repulsion_shift; /**< VdW shift repulsion constants */
192 switch_consts_t vdw_switch; /**< VdW switch constants */
194 /* LJ non-bonded parameters - accessed through texture memory */
195 float *nbfp; /**< nonbonded parameter table with C6/C12 pairs per atom type-pair, 2*ntype^2 elements */
196 cudaTextureObject_t nbfp_texobj; /**< texture object bound to nbfp */
197 float *nbfp_comb; /**< nonbonded parameter table per atom type, 2*ntype elements */
198 cudaTextureObject_t nbfp_comb_texobj; /**< texture object bound to nbfp_texobj */
200 /* Ewald Coulomb force table data - accessed through texture memory */
201 float coulomb_tab_scale; /**< table scale/spacing */
202 float *coulomb_tab; /**< pointer to the table in the device memory */
203 cudaTextureObject_t coulomb_tab_texobj; /**< texture object bound to coulomb_tab */
207 * \brief Pair list data.
211 int na_c; /**< number of atoms per cluster */
213 int nsci; /**< size of sci, # of i clusters in the list */
214 int sci_nalloc; /**< allocation size of sci */
215 nbnxn_sci_t *sci; /**< list of i-cluster ("super-clusters") */
217 int ncj4; /**< total # of 4*j clusters */
218 int cj4_nalloc; /**< allocation size of cj4 */
219 nbnxn_cj4_t *cj4; /**< 4*j cluster list, contains j cluster number
220 and index into the i cluster list */
221 int nimask; /**< # of 4*j clusters * # of warps */
222 int imask_nalloc; /**< allocation size of imask */
223 unsigned int *imask; /**< imask for 2 warps for each 4*j cluster group */
224 nbnxn_excl_t *excl; /**< atom interaction bits */
225 int nexcl; /**< count for excl */
226 int excl_nalloc; /**< allocation size of excl */
228 /* parameter+variables for normal and rolling pruning */
229 bool haveFreshList; /**< true after search, indictes that initial pruning with outer prunning is needed */
230 int rollingPruningNumParts; /**< the number of parts/steps over which one cyle of roling pruning takes places */
231 int rollingPruningPart; /**< the next part to which the roling pruning needs to be applied */
235 * \brief CUDA events used for timing GPU kernels and H2D/D2H transfers.
237 * The two-sized arrays hold the local and non-local values and should always
238 * be indexed with eintLocal/eintNonlocal.
242 cudaEvent_t start_atdat; /**< start event for atom data transfer (every PS step) */
243 cudaEvent_t stop_atdat; /**< stop event for atom data transfer (every PS step) */
244 cudaEvent_t start_nb_h2d[2]; /**< start events for x/q H2D transfers (l/nl, every step) */
245 cudaEvent_t stop_nb_h2d[2]; /**< stop events for x/q H2D transfers (l/nl, every step) */
246 cudaEvent_t start_nb_d2h[2]; /**< start events for f D2H transfer (l/nl, every step) */
247 cudaEvent_t stop_nb_d2h[2]; /**< stop events for f D2H transfer (l/nl, every step) */
248 cudaEvent_t start_pl_h2d[2]; /**< start events for pair-list H2D transfers (l/nl, every PS step) */
249 cudaEvent_t stop_pl_h2d[2]; /**< start events for pair-list H2D transfers (l/nl, every PS step) */
250 bool didPairlistH2D[2]; /**< true when a pair-list transfer has been done at this step */
251 cudaEvent_t start_nb_k[2]; /**< start event for non-bonded kernels (l/nl, every step) */
252 cudaEvent_t stop_nb_k[2]; /**< stop event non-bonded kernels (l/nl, every step) */
253 cudaEvent_t start_prune_k[2]; /**< start event for the 1st pass list pruning kernel (l/nl, every PS step) */
254 cudaEvent_t stop_prune_k[2]; /**< stop event for the 1st pass list pruning kernel (l/nl, every PS step) */
255 bool didPrune[2]; /**< true when we timed pruning and the timings need to be accounted for */
256 cudaEvent_t start_rollingPrune_k[2]; /**< start event for rolling pruning kernels (l/nl, frequency depends on chunk size) */
257 cudaEvent_t stop_rollingPrune_k[2]; /**< stop event for rolling pruning kernels (l/nl, frequency depends on chunk size) */
258 bool didRollingPrune[2]; /**< true when we timed rolling pruning (at the previous step) and the timings need to be accounted for */
262 * \brief Main data structure for CUDA nonbonded force calculations.
264 struct gmx_nbnxn_cuda_t
266 const gmx_device_info_t *dev_info; /**< CUDA device information */
267 bool bUseTwoStreams; /**< true if doing both local/non-local NB work on GPU */
268 cu_atomdata_t *atdat; /**< atom data */
269 cu_nbparam_t *nbparam; /**< parameters required for the non-bonded calc. */
270 cu_plist_t *plist[2]; /**< pair-list data structures (local and non-local) */
271 nb_staging_t nbst; /**< staging area where fshift/energies get downloaded */
273 cudaStream_t stream[2]; /**< local and non-local GPU streams */
275 /** events used for synchronization */
276 cudaEvent_t nonlocal_done; /**< event triggered when the non-local non-bonded kernel
277 is done (and the local transfer can proceed) */
278 cudaEvent_t misc_ops_and_local_H2D_done; /**< event triggered when the tasks issued in
279 the local stream that need to precede the
280 non-local force calculations are done
281 (e.g. f buffer 0-ing, local x/q H2D) */
283 /* NOTE: With current CUDA versions (<=5.0) timing doesn't work with multiple
284 * concurrent streams, so we won't time if both l/nl work is done on GPUs.
285 * Timer init/uninit is still done even with timing off so only the condition
286 * setting bDoTime needs to be change if this CUDA "feature" gets fixed. */
287 bool bDoTime; /**< True if event-based timing is enabled. */
288 cu_timers_t *timers; /**< CUDA event-based timers. */
289 gmx_wallclock_gpu_t *timings; /**< Timing data. */
296 #endif /* NBNXN_CUDA_TYPES_H */