2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014,2015,2016,2017,2018,2019, 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 Define CUDA implementation of nbnxn_gpu.h
38 * \author Szilard Pall <pall.szilard@gmail.com>
47 #include "gromacs/mdlib/nbnxn_gpu.h"
54 #include "nbnxn_cuda.h"
56 #include "gromacs/gpu_utils/cudautils.cuh"
57 #include "gromacs/mdlib/force_flags.h"
58 #include "gromacs/mdlib/nb_verlet.h"
59 #include "gromacs/mdlib/nbnxn_gpu_common.h"
60 #include "gromacs/mdlib/nbnxn_gpu_common_utils.h"
61 #include "gromacs/mdlib/nbnxn_gpu_data_mgmt.h"
62 #include "gromacs/mdlib/nbnxn_pairlist.h"
63 #include "gromacs/timing/gpu_timing.h"
64 #include "gromacs/utility/cstringutil.h"
65 #include "gromacs/utility/gmxassert.h"
67 #include "nbnxn_cuda_types.h"
70 /***** The kernel declarations/definitions come here *****/
72 /* Top-level kernel declaration generation: will generate through multiple
73 * inclusion the following flavors for all kernel declarations:
74 * - force-only output;
75 * - force and energy output;
76 * - force-only with pair list pruning;
77 * - force and energy output with pair list pruning.
79 #define FUNCTION_DECLARATION_ONLY
81 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernels.cuh"
82 /** Force & energy **/
84 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernels.cuh"
87 /*** Pair-list pruning kernels ***/
90 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernels.cuh"
91 /** Force & energy **/
93 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernels.cuh"
97 /* Prune-only kernels */
98 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_pruneonly.cuh"
99 #undef FUNCTION_DECLARATION_ONLY
101 /* Now generate the function definitions if we are using a single compilation unit. */
102 #if GMX_CUDA_NB_SINGLE_COMPILATION_UNIT
103 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_F_noprune.cu"
104 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_F_prune.cu"
105 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_VF_noprune.cu"
106 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_VF_prune.cu"
107 #include "gromacs/mdlib/nbnxn_cuda/nbnxn_cuda_kernel_pruneonly.cu"
108 #endif /* GMX_CUDA_NB_SINGLE_COMPILATION_UNIT */
111 /*! Nonbonded kernel function pointer type */
112 typedef void (*nbnxn_cu_kfunc_ptr_t)(const cu_atomdata_t,
117 /*********************************/
119 /*! Returns the number of blocks to be used for the nonbonded GPU kernel. */
120 static inline int calc_nb_kernel_nblock(int nwork_units, const gmx_device_info_t *dinfo)
125 /* CUDA does not accept grid dimension of 0 (which can happen e.g. with an
126 empty domain) and that case should be handled before this point. */
127 assert(nwork_units > 0);
129 max_grid_x_size = dinfo->prop.maxGridSize[0];
131 /* do we exceed the grid x dimension limit? */
132 if (nwork_units > max_grid_x_size)
134 gmx_fatal(FARGS, "Watch out, the input system is too large to simulate!\n"
135 "The number of nonbonded work units (=number of super-clusters) exceeds the"
136 "maximum grid size in x dimension (%d > %d)!", nwork_units, max_grid_x_size);
143 /* Constant arrays listing all kernel function pointers and enabling selection
144 of a kernel in an elegant manner. */
146 /*! Pointers to the non-bonded kernels organized in 2-dim arrays by:
147 * electrostatics and VDW type.
149 * Note that the row- and column-order of function pointers has to match the
150 * order of corresponding enumerated electrostatics and vdw types, resp.,
151 * defined in nbnxn_cuda_types.h.
154 /*! Force-only kernel function pointers. */
155 static const nbnxn_cu_kfunc_ptr_t nb_kfunc_noener_noprune_ptr[eelCuNR][evdwCuNR] =
157 { nbnxn_kernel_ElecCut_VdwLJ_F_cuda, nbnxn_kernel_ElecCut_VdwLJCombGeom_F_cuda, nbnxn_kernel_ElecCut_VdwLJCombLB_F_cuda, nbnxn_kernel_ElecCut_VdwLJFsw_F_cuda, nbnxn_kernel_ElecCut_VdwLJPsw_F_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombGeom_F_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombLB_F_cuda },
158 { nbnxn_kernel_ElecRF_VdwLJ_F_cuda, nbnxn_kernel_ElecRF_VdwLJCombGeom_F_cuda, nbnxn_kernel_ElecRF_VdwLJCombLB_F_cuda, nbnxn_kernel_ElecRF_VdwLJFsw_F_cuda, nbnxn_kernel_ElecRF_VdwLJPsw_F_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombGeom_F_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombLB_F_cuda },
159 { nbnxn_kernel_ElecEwQSTab_VdwLJ_F_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombGeom_F_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombLB_F_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJFsw_F_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJPsw_F_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombGeom_F_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombLB_F_cuda },
160 { nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJ_F_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombGeom_F_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombLB_F_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJFsw_F_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJPsw_F_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombGeom_F_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombLB_F_cuda },
161 { nbnxn_kernel_ElecEw_VdwLJ_F_cuda, nbnxn_kernel_ElecEw_VdwLJCombGeom_F_cuda, nbnxn_kernel_ElecEw_VdwLJCombLB_F_cuda, nbnxn_kernel_ElecEw_VdwLJFsw_F_cuda, nbnxn_kernel_ElecEw_VdwLJPsw_F_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombGeom_F_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombLB_F_cuda },
162 { nbnxn_kernel_ElecEwTwinCut_VdwLJ_F_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_F_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_F_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJFsw_F_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJPsw_F_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_F_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombLB_F_cuda }
165 /*! Force + energy kernel function pointers. */
166 static const nbnxn_cu_kfunc_ptr_t nb_kfunc_ener_noprune_ptr[eelCuNR][evdwCuNR] =
168 { nbnxn_kernel_ElecCut_VdwLJ_VF_cuda, nbnxn_kernel_ElecCut_VdwLJCombGeom_VF_cuda, nbnxn_kernel_ElecCut_VdwLJCombLB_VF_cuda, nbnxn_kernel_ElecCut_VdwLJFsw_VF_cuda, nbnxn_kernel_ElecCut_VdwLJPsw_VF_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombGeom_VF_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombLB_VF_cuda },
169 { nbnxn_kernel_ElecRF_VdwLJ_VF_cuda, nbnxn_kernel_ElecRF_VdwLJCombGeom_VF_cuda, nbnxn_kernel_ElecRF_VdwLJCombLB_VF_cuda, nbnxn_kernel_ElecRF_VdwLJFsw_VF_cuda, nbnxn_kernel_ElecRF_VdwLJPsw_VF_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombGeom_VF_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombLB_VF_cuda },
170 { nbnxn_kernel_ElecEwQSTab_VdwLJ_VF_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombGeom_VF_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombLB_VF_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJFsw_VF_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJPsw_VF_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombGeom_VF_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombLB_VF_cuda },
171 { nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJ_VF_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombGeom_VF_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombLB_VF_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJFsw_VF_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJPsw_VF_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombGeom_VF_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombLB_VF_cuda },
172 { nbnxn_kernel_ElecEw_VdwLJ_VF_cuda, nbnxn_kernel_ElecEw_VdwLJCombGeom_VF_cuda, nbnxn_kernel_ElecEw_VdwLJCombLB_VF_cuda, nbnxn_kernel_ElecEw_VdwLJFsw_VF_cuda, nbnxn_kernel_ElecEw_VdwLJPsw_VF_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombGeom_VF_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombLB_VF_cuda },
173 { nbnxn_kernel_ElecEwTwinCut_VdwLJ_VF_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_VF_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_VF_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJFsw_VF_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJPsw_VF_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_VF_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombLB_VF_cuda }
176 /*! Force + pruning kernel function pointers. */
177 static const nbnxn_cu_kfunc_ptr_t nb_kfunc_noener_prune_ptr[eelCuNR][evdwCuNR] =
179 { nbnxn_kernel_ElecCut_VdwLJ_F_prune_cuda, nbnxn_kernel_ElecCut_VdwLJCombGeom_F_prune_cuda, nbnxn_kernel_ElecCut_VdwLJCombLB_F_prune_cuda, nbnxn_kernel_ElecCut_VdwLJFsw_F_prune_cuda, nbnxn_kernel_ElecCut_VdwLJPsw_F_prune_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombGeom_F_prune_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombLB_F_prune_cuda },
180 { nbnxn_kernel_ElecRF_VdwLJ_F_prune_cuda, nbnxn_kernel_ElecRF_VdwLJCombGeom_F_prune_cuda, nbnxn_kernel_ElecRF_VdwLJCombLB_F_prune_cuda, nbnxn_kernel_ElecRF_VdwLJFsw_F_prune_cuda, nbnxn_kernel_ElecRF_VdwLJPsw_F_prune_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombGeom_F_prune_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombLB_F_prune_cuda },
181 { nbnxn_kernel_ElecEwQSTab_VdwLJ_F_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombGeom_F_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombLB_F_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJFsw_F_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJPsw_F_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombGeom_F_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombLB_F_prune_cuda },
182 { nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJ_F_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombGeom_F_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombLB_F_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJFsw_F_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJPsw_F_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombGeom_F_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombLB_F_prune_cuda },
183 { nbnxn_kernel_ElecEw_VdwLJ_F_prune_cuda, nbnxn_kernel_ElecEw_VdwLJCombGeom_F_prune_cuda, nbnxn_kernel_ElecEw_VdwLJCombLB_F_prune_cuda, nbnxn_kernel_ElecEw_VdwLJFsw_F_prune_cuda, nbnxn_kernel_ElecEw_VdwLJPsw_F_prune_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombGeom_F_prune_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombLB_F_prune_cuda },
184 { nbnxn_kernel_ElecEwTwinCut_VdwLJ_F_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_F_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_F_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJFsw_F_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJPsw_F_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_F_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombLB_F_prune_cuda }
187 /*! Force + energy + pruning kernel function pointers. */
188 static const nbnxn_cu_kfunc_ptr_t nb_kfunc_ener_prune_ptr[eelCuNR][evdwCuNR] =
190 { nbnxn_kernel_ElecCut_VdwLJ_VF_prune_cuda, nbnxn_kernel_ElecCut_VdwLJCombGeom_VF_prune_cuda, nbnxn_kernel_ElecCut_VdwLJCombLB_VF_prune_cuda, nbnxn_kernel_ElecCut_VdwLJFsw_VF_prune_cuda, nbnxn_kernel_ElecCut_VdwLJPsw_VF_prune_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombGeom_VF_prune_cuda, nbnxn_kernel_ElecCut_VdwLJEwCombLB_VF_prune_cuda },
191 { nbnxn_kernel_ElecRF_VdwLJ_VF_prune_cuda, nbnxn_kernel_ElecRF_VdwLJCombGeom_VF_prune_cuda, nbnxn_kernel_ElecRF_VdwLJCombLB_VF_prune_cuda, nbnxn_kernel_ElecRF_VdwLJFsw_VF_prune_cuda, nbnxn_kernel_ElecRF_VdwLJPsw_VF_prune_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombGeom_VF_prune_cuda, nbnxn_kernel_ElecRF_VdwLJEwCombLB_VF_prune_cuda },
192 { nbnxn_kernel_ElecEwQSTab_VdwLJ_VF_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJCombLB_VF_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJFsw_VF_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJPsw_VF_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEwQSTab_VdwLJEwCombLB_VF_prune_cuda },
193 { nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJ_VF_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJCombLB_VF_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJFsw_VF_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJPsw_VF_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEwQSTabTwinCut_VdwLJEwCombLB_VF_prune_cuda },
194 { nbnxn_kernel_ElecEw_VdwLJ_VF_prune_cuda, nbnxn_kernel_ElecEw_VdwLJCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEw_VdwLJCombLB_VF_prune_cuda, nbnxn_kernel_ElecEw_VdwLJFsw_VF_prune_cuda, nbnxn_kernel_ElecEw_VdwLJPsw_VF_prune_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEw_VdwLJEwCombLB_VF_prune_cuda },
195 { nbnxn_kernel_ElecEwTwinCut_VdwLJ_VF_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJCombLB_VF_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJFsw_VF_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJPsw_VF_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombGeom_VF_prune_cuda, nbnxn_kernel_ElecEwTwinCut_VdwLJEwCombLB_VF_prune_cuda }
198 /*! Return a pointer to the kernel version to be executed at the current step. */
199 static inline nbnxn_cu_kfunc_ptr_t select_nbnxn_kernel(int eeltype,
203 const gmx_device_info_t gmx_unused *devInfo)
205 nbnxn_cu_kfunc_ptr_t res;
207 GMX_ASSERT(eeltype < eelCuNR,
208 "The electrostatics type requested is not implemented in the CUDA kernels.");
209 GMX_ASSERT(evdwtype < evdwCuNR,
210 "The VdW type requested is not implemented in the CUDA kernels.");
212 /* assert assumptions made by the kernels */
213 GMX_ASSERT(c_nbnxnGpuClusterSize*c_nbnxnGpuClusterSize/c_nbnxnGpuClusterpairSplit == devInfo->prop.warpSize,
214 "The CUDA kernels require the cluster_size_i*cluster_size_j/nbnxn_gpu_clusterpair_split to match the warp size of the architecture targeted.");
220 res = nb_kfunc_ener_prune_ptr[eeltype][evdwtype];
224 res = nb_kfunc_ener_noprune_ptr[eeltype][evdwtype];
231 res = nb_kfunc_noener_prune_ptr[eeltype][evdwtype];
235 res = nb_kfunc_noener_noprune_ptr[eeltype][evdwtype];
242 /*! \brief Calculates the amount of shared memory required by the nonbonded kernel in use. */
243 static inline int calc_shmem_required_nonbonded(const int num_threads_z, const gmx_device_info_t gmx_unused *dinfo, const cu_nbparam_t *nbp)
249 /* size of shmem (force-buffers/xq/atom type preloading) */
250 /* NOTE: with the default kernel on sm3.0 we need shmem only for pre-loading */
251 /* i-atom x+q in shared memory */
252 shmem = c_numClPerSupercl * c_clSize * sizeof(float4);
253 /* cj in shared memory, for each warp separately */
254 shmem += num_threads_z * c_nbnxnGpuClusterpairSplit * c_nbnxnGpuJgroupSize * sizeof(int);
256 if (nbp->vdwtype == evdwCuCUTCOMBGEOM ||
257 nbp->vdwtype == evdwCuCUTCOMBLB)
259 /* i-atom LJ combination parameters in shared memory */
260 shmem += c_numClPerSupercl * c_clSize * sizeof(float2);
264 /* i-atom types in shared memory */
265 shmem += c_numClPerSupercl * c_clSize * sizeof(int);
271 /*! \brief Launch asynchronously the xq buffer host to device copy. */
272 void nbnxn_gpu_copy_xq_to_gpu(gmx_nbnxn_cuda_t *nb,
273 const nbnxn_atomdata_t *nbatom,
277 int adat_begin, adat_len; /* local/nonlocal offset and length used for xq and f */
279 cu_atomdata_t *adat = nb->atdat;
280 cu_plist_t *plist = nb->plist[iloc];
281 cu_timers_t *t = nb->timers;
282 cudaStream_t stream = nb->stream[iloc];
284 bool bDoTime = nb->bDoTime;
286 /* Don't launch the non-local H2D copy if there is no dependent
287 work to do: neither non-local nor other (e.g. bonded) work
288 to do that has as input the nbnxn coordaintes.
289 Doing the same for the local kernel is more complicated, since the
290 local part of the force array also depends on the non-local kernel.
291 So to avoid complicating the code and to reduce the risk of bugs,
292 we always call the local local x+q copy (and the rest of the local
293 work in nbnxn_gpu_launch_kernel().
295 if (!haveOtherWork && canSkipWork(nb, iloc))
297 plist->haveFreshList = false;
302 /* calculate the atom data index range based on locality */
306 adat_len = adat->natoms_local;
310 adat_begin = adat->natoms_local;
311 adat_len = adat->natoms - adat->natoms_local;
314 /* beginning of timed HtoD section */
317 t->nb_h2d[iloc].openTimingRegion(stream);
321 cu_copy_H2D_async(adat->xq + adat_begin,
322 static_cast<const void *>(nbatom->x().data() + adat_begin * 4),
323 adat_len * sizeof(*adat->xq), stream);
327 t->nb_h2d[iloc].closeTimingRegion(stream);
330 /* When we get here all misc operations issued in the local stream as well as
331 the local xq H2D are done,
332 so we record that in the local stream and wait for it in the nonlocal one.
333 This wait needs to precede any PP tasks, bonded or nonbonded, that may
334 compute on interactions between local and nonlocal atoms.
336 if (nb->bUseTwoStreams)
338 if (iloc == eintLocal)
340 cudaError_t stat = cudaEventRecord(nb->misc_ops_and_local_H2D_done, stream);
341 CU_RET_ERR(stat, "cudaEventRecord on misc_ops_and_local_H2D_done failed");
345 cudaError_t stat = cudaStreamWaitEvent(stream, nb->misc_ops_and_local_H2D_done, 0);
346 CU_RET_ERR(stat, "cudaStreamWaitEvent on misc_ops_and_local_H2D_done failed");
351 /*! As we execute nonbonded workload in separate streams, before launching
352 the kernel we need to make sure that he following operations have completed:
353 - atomdata allocation and related H2D transfers (every nstlist step);
354 - pair list H2D transfer (every nstlist step);
355 - shift vector H2D transfer (every nstlist step);
356 - force (+shift force and energy) output clearing (every step).
358 These operations are issued in the local stream at the beginning of the step
359 and therefore always complete before the local kernel launch. The non-local
360 kernel is launched after the local on the same device/context hence it is
361 inherently scheduled after the operations in the local stream (including the
362 above "misc_ops") on pre-GK110 devices with single hardware queue, but on later
363 devices with multiple hardware queues the dependency needs to be enforced.
364 We use the misc_ops_and_local_H2D_done event to record the point where
365 the local x+q H2D (and all preceding) tasks are complete and synchronize
366 with this event in the non-local stream before launching the non-bonded kernel.
368 void nbnxn_gpu_launch_kernel(gmx_nbnxn_cuda_t *nb,
372 /* CUDA kernel launch-related stuff */
374 dim3 dim_block, dim_grid;
375 nbnxn_cu_kfunc_ptr_t nb_kernel = nullptr; /* fn pointer to the nonbonded kernel */
377 cu_atomdata_t *adat = nb->atdat;
378 cu_nbparam_t *nbp = nb->nbparam;
379 cu_plist_t *plist = nb->plist[iloc];
380 cu_timers_t *t = nb->timers;
381 cudaStream_t stream = nb->stream[iloc];
383 bool bCalcEner = flags & GMX_FORCE_ENERGY;
384 bool bCalcFshift = flags & GMX_FORCE_VIRIAL;
385 bool bDoTime = nb->bDoTime;
387 /* Don't launch the non-local kernel if there is no work to do.
388 Doing the same for the local kernel is more complicated, since the
389 local part of the force array also depends on the non-local kernel.
390 So to avoid complicating the code and to reduce the risk of bugs,
391 we always call the local kernel, and later (not in
392 this function) the stream wait, local f copyback and the f buffer
393 clearing. All these operations, except for the local interaction kernel,
394 are needed for the non-local interactions. The skip of the local kernel
395 call is taken care of later in this function. */
396 if (canSkipWork(nb, iloc))
398 plist->haveFreshList = false;
403 if (nbp->useDynamicPruning && plist->haveFreshList)
405 /* Prunes for rlistOuter and rlistInner, sets plist->haveFreshList=false
406 (TODO: ATM that's the way the timing accounting can distinguish between
407 separate prune kernel and combined force+prune, maybe we need a better way?).
409 nbnxn_gpu_launch_kernel_pruneonly(nb, iloc, 1);
412 if (plist->nsci == 0)
414 /* Don't launch an empty local kernel (not allowed with CUDA) */
418 /* beginning of timed nonbonded calculation section */
421 t->nb_k[iloc].openTimingRegion(stream);
424 /* get the pointer to the kernel flavor we need to use */
425 nb_kernel = select_nbnxn_kernel(nbp->eeltype,
428 (plist->haveFreshList && !nb->timers->didPrune[iloc]),
431 /* Kernel launch config:
432 * - The thread block dimensions match the size of i-clusters, j-clusters,
433 * and j-cluster concurrency, in x, y, and z, respectively.
434 * - The 1D block-grid contains as many blocks as super-clusters.
436 int num_threads_z = 1;
437 if (nb->dev_info->prop.major == 3 && nb->dev_info->prop.minor == 7)
441 nblock = calc_nb_kernel_nblock(plist->nsci, nb->dev_info);
443 KernelLaunchConfig config;
444 config.blockSize[0] = c_clSize;
445 config.blockSize[1] = c_clSize;
446 config.blockSize[2] = num_threads_z;
447 config.gridSize[0] = nblock;
448 config.sharedMemorySize = calc_shmem_required_nonbonded(num_threads_z, nb->dev_info, nbp);
449 config.stream = stream;
453 fprintf(debug, "Non-bonded GPU launch configuration:\n\tThread block: %zux%zux%zu\n\t"
454 "\tGrid: %zux%zu\n\t#Super-clusters/clusters: %d/%d (%d)\n"
456 config.blockSize[0], config.blockSize[1], config.blockSize[2],
457 config.gridSize[0], config.gridSize[1], plist->nsci*c_numClPerSupercl,
458 c_numClPerSupercl, plist->na_c,
459 config.sharedMemorySize);
462 auto *timingEvent = bDoTime ? t->nb_k[iloc].fetchNextEvent() : nullptr;
463 const auto kernelArgs = prepareGpuKernelArguments(nb_kernel, config, adat, nbp, plist, &bCalcFshift);
464 launchGpuKernel(nb_kernel, config, timingEvent, "k_calc_nb", kernelArgs);
468 t->nb_k[iloc].closeTimingRegion(stream);
471 if (GMX_NATIVE_WINDOWS)
473 /* Windows: force flushing WDDM queue */
474 cudaStreamQuery(stream);
478 /*! Calculates the amount of shared memory required by the CUDA kernel in use. */
479 static inline int calc_shmem_required_prune(const int num_threads_z)
483 /* i-atom x in shared memory */
484 shmem = c_numClPerSupercl * c_clSize * sizeof(float4);
485 /* cj in shared memory, for each warp separately */
486 shmem += num_threads_z * c_nbnxnGpuClusterpairSplit * c_nbnxnGpuJgroupSize * sizeof(int);
491 void nbnxn_gpu_launch_kernel_pruneonly(gmx_nbnxn_cuda_t *nb,
495 cu_atomdata_t *adat = nb->atdat;
496 cu_nbparam_t *nbp = nb->nbparam;
497 cu_plist_t *plist = nb->plist[iloc];
498 cu_timers_t *t = nb->timers;
499 cudaStream_t stream = nb->stream[iloc];
501 bool bDoTime = nb->bDoTime;
503 if (plist->haveFreshList)
505 GMX_ASSERT(numParts == 1, "With first pruning we expect 1 part");
507 /* Set rollingPruningNumParts to signal that it is not set */
508 plist->rollingPruningNumParts = 0;
509 plist->rollingPruningPart = 0;
513 if (plist->rollingPruningNumParts == 0)
515 plist->rollingPruningNumParts = numParts;
519 GMX_ASSERT(numParts == plist->rollingPruningNumParts, "It is not allowed to change numParts in between list generation steps");
523 /* Use a local variable for part and update in plist, so we can return here
524 * without duplicating the part increment code.
526 int part = plist->rollingPruningPart;
528 plist->rollingPruningPart++;
529 if (plist->rollingPruningPart >= plist->rollingPruningNumParts)
531 plist->rollingPruningPart = 0;
534 /* Compute the number of list entries to prune in this pass */
535 int numSciInPart = (plist->nsci - part)/numParts;
537 /* Don't launch the kernel if there is no work to do (not allowed with CUDA) */
538 if (numSciInPart <= 0)
540 plist->haveFreshList = false;
545 GpuRegionTimer *timer = nullptr;
548 timer = &(plist->haveFreshList ? t->prune_k[iloc] : t->rollingPrune_k[iloc]);
551 /* beginning of timed prune calculation section */
554 timer->openTimingRegion(stream);
557 /* Kernel launch config:
558 * - The thread block dimensions match the size of i-clusters, j-clusters,
559 * and j-cluster concurrency, in x, y, and z, respectively.
560 * - The 1D block-grid contains as many blocks as super-clusters.
562 int num_threads_z = c_cudaPruneKernelJ4Concurrency;
563 int nblock = calc_nb_kernel_nblock(numSciInPart, nb->dev_info);
564 KernelLaunchConfig config;
565 config.blockSize[0] = c_clSize;
566 config.blockSize[1] = c_clSize;
567 config.blockSize[2] = num_threads_z;
568 config.gridSize[0] = nblock;
569 config.sharedMemorySize = calc_shmem_required_prune(num_threads_z);
570 config.stream = stream;
574 fprintf(debug, "Pruning GPU kernel launch configuration:\n\tThread block: %zux%zux%zu\n\t"
575 "\tGrid: %zux%zu\n\t#Super-clusters/clusters: %d/%d (%d)\n"
577 config.blockSize[0], config.blockSize[1], config.blockSize[2],
578 config.gridSize[0], config.gridSize[1], numSciInPart*c_numClPerSupercl,
579 c_numClPerSupercl, plist->na_c,
580 config.sharedMemorySize);
583 auto *timingEvent = bDoTime ? timer->fetchNextEvent() : nullptr;
584 constexpr char kernelName[] = "k_pruneonly";
585 const auto &kernel = plist->haveFreshList ? nbnxn_kernel_prune_cuda<true> : nbnxn_kernel_prune_cuda<false>;
586 const auto kernelArgs = prepareGpuKernelArguments(kernel, config, adat, nbp, plist, &numParts, &part);
587 launchGpuKernel(kernel, config, timingEvent, kernelName, kernelArgs);
589 /* TODO: consider a more elegant way to track which kernel has been called
590 (combined or separate 1st pass prune, rolling prune). */
591 if (plist->haveFreshList)
593 plist->haveFreshList = false;
594 /* Mark that pruning has been done */
595 nb->timers->didPrune[iloc] = true;
599 /* Mark that rolling pruning has been done */
600 nb->timers->didRollingPrune[iloc] = true;
605 timer->closeTimingRegion(stream);
608 if (GMX_NATIVE_WINDOWS)
610 /* Windows: force flushing WDDM queue */
611 cudaStreamQuery(stream);
615 void nbnxn_gpu_launch_cpyback(gmx_nbnxn_cuda_t *nb,
616 nbnxn_atomdata_t *nbatom,
622 int adat_begin, adat_len; /* local/nonlocal offset and length used for xq and f */
624 /* determine interaction locality from atom locality */
625 int iloc = gpuAtomToInteractionLocality(aloc);
627 cu_atomdata_t *adat = nb->atdat;
628 cu_timers_t *t = nb->timers;
629 bool bDoTime = nb->bDoTime;
630 cudaStream_t stream = nb->stream[iloc];
632 bool bCalcEner = flags & GMX_FORCE_ENERGY;
633 bool bCalcFshift = flags & GMX_FORCE_VIRIAL;
635 /* don't launch non-local copy-back if there was no non-local work to do */
636 if (!haveOtherWork && canSkipWork(nb, iloc))
641 getGpuAtomRange(adat, aloc, &adat_begin, &adat_len);
643 /* beginning of timed D2H section */
646 t->nb_d2h[iloc].openTimingRegion(stream);
649 /* With DD the local D2H transfer can only start after the non-local
650 kernel has finished. */
651 if (iloc == eintLocal && nb->bUseTwoStreams)
653 stat = cudaStreamWaitEvent(stream, nb->nonlocal_done, 0);
654 CU_RET_ERR(stat, "cudaStreamWaitEvent on nonlocal_done failed");
658 cu_copy_D2H_async(nbatom->out[0].f.data() + adat_begin * 3, adat->f + adat_begin,
659 (adat_len)*sizeof(*adat->f), stream);
661 /* After the non-local D2H is launched the nonlocal_done event can be
662 recorded which signals that the local D2H can proceed. This event is not
663 placed after the non-local kernel because we want the non-local data
665 if (iloc == eintNonlocal)
667 stat = cudaEventRecord(nb->nonlocal_done, stream);
668 CU_RET_ERR(stat, "cudaEventRecord on nonlocal_done failed");
671 /* only transfer energies in the local stream */
677 cu_copy_D2H_async(nb->nbst.fshift, adat->fshift,
678 SHIFTS * sizeof(*nb->nbst.fshift), stream);
684 cu_copy_D2H_async(nb->nbst.e_lj, adat->e_lj,
685 sizeof(*nb->nbst.e_lj), stream);
686 cu_copy_D2H_async(nb->nbst.e_el, adat->e_el,
687 sizeof(*nb->nbst.e_el), stream);
693 t->nb_d2h[iloc].closeTimingRegion(stream);
697 void nbnxn_cuda_set_cacheconfig()
701 for (int i = 0; i < eelCuNR; i++)
703 for (int j = 0; j < evdwCuNR; j++)
705 /* Default kernel 32/32 kB Shared/L1 */
706 cudaFuncSetCacheConfig(nb_kfunc_ener_prune_ptr[i][j], cudaFuncCachePreferEqual);
707 cudaFuncSetCacheConfig(nb_kfunc_ener_noprune_ptr[i][j], cudaFuncCachePreferEqual);
708 cudaFuncSetCacheConfig(nb_kfunc_noener_prune_ptr[i][j], cudaFuncCachePreferEqual);
709 stat = cudaFuncSetCacheConfig(nb_kfunc_noener_noprune_ptr[i][j], cudaFuncCachePreferEqual);
710 CU_RET_ERR(stat, "cudaFuncSetCacheConfig failed");