2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2018,2019,2020, 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.
37 * Declares PmeGpuProgramImpl, which stores PME GPU (compiled) kernel handles.
39 * \author Aleksei Iupinov <a.yupinov@gmail.com>
40 * \ingroup module_ewald
42 #ifndef GMX_EWALD_PME_PME_GPU_PROGRAM_IMPL_H
43 #define GMX_EWALD_PME_PME_GPU_PROGRAM_IMPL_H
47 #include "gromacs/gpu_utils/device_context.h"
48 #include "gromacs/gpu_utils/gputraits.h"
49 #include "gromacs/utility/classhelpers.h"
51 struct DeviceInformation;
55 * PME GPU persistent host program/kernel data, which should be initialized once for the whole execution.
57 * Primary purpose of this is to not recompile GPU kernels for each OpenCL unit test,
58 * while the relevant GPU context (e.g. cl_context) instance persists.
59 * In CUDA, this just assigns the kernel function pointers.
60 * This also implicitly relies on the fact that reasonable share of the kernels are always used.
61 * If there were more template parameters, even smaller share of all possible kernels would be used.
63 * \todo In future if we would need to react to either user input or
64 * auto-tuning to compile different kernels, then we might wish to
65 * revisit the number of kernels we pre-compile, and/or the management
68 * This also doesn't manage cuFFT/clFFT kernels, which depend on the PME grid dimensions.
70 * TODO: pass cl_context to the constructor and not create it inside.
71 * See also Redmine #2522.
73 struct PmeGpuProgramImpl
76 * This is a handle to the GPU context, which is just a dummy in CUDA,
77 * but is created/destroyed by this class in OpenCL.
78 * TODO: Later we want to be able to own the context at a higher level and not here,
79 * but this class would still need the non-owning context handle to build the kernels.
81 DeviceContext deviceContext_;
83 //! Conveniently all the PME kernels use the same single argument type
84 #if GMX_GPU == GMX_GPU_CUDA
85 using PmeKernelHandle = void (*)(const struct PmeGpuCudaKernelParams);
86 #elif GMX_GPU == GMX_GPU_OPENCL
87 using PmeKernelHandle = cl_kernel;
89 using PmeKernelHandle = void*;
93 * Maximum synchronous GPU thread group execution width.
94 * "Warp" is a CUDA term which we end up reusing in OpenCL kernels as well.
95 * For CUDA, this is a static value that comes from gromacs/gpu_utils/cuda_arch_utils.cuh;
96 * for OpenCL, we have to query it dynamically.
102 * Spread/spline kernels are compiled only for order of 4.
103 * There are multiple versions of each kernel, paramaretized according to
104 * Number of threads per atom. Using either order(4) or order*order (16) threads per atom is
105 * supported If the spline data is written in the spline/spread kernel and loaded in the gather
106 * or recalculated in the gather.
107 * Spreading kernels also have hardcoded X/Y indices wrapping parameters,
108 * as a placeholder for implementing 1/2D decomposition.
110 size_t spreadWorkGroupSize;
112 PmeKernelHandle splineKernel;
113 PmeKernelHandle splineKernelThPerAtom4;
114 PmeKernelHandle spreadKernel;
115 PmeKernelHandle spreadKernelThPerAtom4;
116 PmeKernelHandle splineAndSpreadKernel;
117 PmeKernelHandle splineAndSpreadKernelThPerAtom4;
118 PmeKernelHandle splineAndSpreadKernelWriteSplines;
119 PmeKernelHandle splineAndSpreadKernelWriteSplinesThPerAtom4;
123 /** Same for gather: hardcoded X/Y unwrap parameters, order of 4, plus
124 * it can either reduce with previous forces in the host buffer, or ignore them.
125 * Also similarly to the gather we can use either order(4) or order*order (16) threads per atom
126 * and either recalculate the splines or read the ones written by the spread
128 size_t gatherWorkGroupSize;
130 PmeKernelHandle gatherKernel;
131 PmeKernelHandle gatherKernelThPerAtom4;
132 PmeKernelHandle gatherKernelReadSplines;
133 PmeKernelHandle gatherKernelReadSplinesThPerAtom4;
137 /** Solve kernel doesn't care about the interpolation order, but can optionally
138 * compute energy and virial, and supports XYZ and YZX grid orderings.
140 size_t solveMaxWorkGroupSize;
142 PmeKernelHandle solveYZXKernel;
143 PmeKernelHandle solveXYZKernel;
144 PmeKernelHandle solveYZXEnergyKernel;
145 PmeKernelHandle solveXYZEnergyKernel;
148 PmeGpuProgramImpl() = delete;
149 //! Constructor for the given device
150 explicit PmeGpuProgramImpl(const DeviceInformation* deviceInfo);
151 ~PmeGpuProgramImpl();
152 GMX_DISALLOW_COPY_AND_ASSIGN(PmeGpuProgramImpl);
155 // Compiles kernels, if supported. Called by the constructor.
156 void compileKernels(const DeviceInformation* deviceInfo);