2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2014,2015,2016,2017 by the GROMACS development team.
5 * Copyright (c) 2018,2019,2020, by the GROMACS development team, led by
6 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
7 * and including many others, as listed in the AUTHORS file in the
8 * top-level source directory and at http://www.gromacs.org.
10 * GROMACS is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public License
12 * as published by the Free Software Foundation; either version 2.1
13 * of the License, or (at your option) any later version.
15 * GROMACS is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with GROMACS; if not, see
22 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
23 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
25 * If you want to redistribute modifications to GROMACS, please
26 * consider that scientific software is very special. Version
27 * control is crucial - bugs must be traceable. We will be happy to
28 * consider code for inclusion in the official distribution, but
29 * derived work must not be called official GROMACS. Details are found
30 * in the README & COPYING files - if they are missing, get the
31 * official version at http://www.gromacs.org.
33 * To help us fund GROMACS development, we humbly ask that you cite
34 * the research papers on the package. Check out http://www.gromacs.org.
36 #ifndef CUDA_ARCH_UTILS_CUH_
37 #define CUDA_ARCH_UTILS_CUH_
39 #include "gromacs/utility/basedefinitions.h"
42 * \brief CUDA arch dependent definitions.
44 * \author Szilard Pall <pall.szilard@gmail.com>
47 /* GMX_PTX_ARCH is set to the virtual arch (PTX) version targeted by
48 * the current compiler pass or zero for the host pass and it is
49 * intended to be used instead of __CUDA_ARCH__.
52 # define GMX_PTX_ARCH 0
54 # define GMX_PTX_ARCH __CUDA_ARCH__
57 /* Until CC 5.2 and likely for the near future all NVIDIA architectures
58 have a warp size of 32, but this could change later. If it does, the
59 following constants should depend on the value of GMX_PTX_ARCH.
61 static const int warp_size = 32;
62 static const int warp_size_log2 = 5;
63 /*! \brief Bitmask corresponding to all threads active in a warp.
64 * NOTE that here too we assume 32-wide warps.
66 static const unsigned int c_fullWarpMask = 0xffffffff;
68 /*! \brief Allow disabling CUDA textures using the GMX_DISABLE_CUDA_TEXTURES macro.
70 * Only texture objects supported.
71 * Disable texture support missing in clang (all versions up to <=5.0-dev as of writing).
73 * This option will not influence functionality. All features using textures ought
74 * to have fallback for texture-less reads (direct/LDG loads), all new code needs
75 * to provide fallback code.
77 #if defined(GMX_DISABLE_CUDA_TEXTURES) || (defined(__clang__) && defined(__CUDA__))
78 # define DISABLE_CUDA_TEXTURES 1
80 # define DISABLE_CUDA_TEXTURES 0
83 /*! \brief True if the use of texture fetch in the CUDA kernels is disabled. */
84 static const bool c_disableCudaTextures = DISABLE_CUDA_TEXTURES;
87 /* CUDA architecture technical characteristics. Needs macros because it is used
88 * in the __launch_bounds__ function qualifiers and might need it in preprocessor
93 # if GMX_PTX_ARCH <= 370 // CC 3.x
94 # define GMX_CUDA_MAX_BLOCKS_PER_MP 16
95 # define GMX_CUDA_MAX_THREADS_PER_MP 2048
97 /* Note that this final branch covers all future architectures (current gen
98 * is 6.x as of writing), hence assuming that these *currently defined* upper
99 * limits will not be lowered.
101 # define GMX_CUDA_MAX_BLOCKS_PER_MP 32
102 # define GMX_CUDA_MAX_THREADS_PER_MP 2048
105 # define GMX_CUDA_MAX_BLOCKS_PER_MP 0
106 # define GMX_CUDA_MAX_THREADS_PER_MP 0
109 // Macro defined for clang CUDA device compilation in the presence of debug symbols
110 // used to work around codegen bug that breaks some kernels when assertions are on
111 // at -O1 and higher (tested with clang 6-8).
112 #if defined(__clang__) && defined(__CUDA__) && defined(__CUDA_ARCH__) && !defined(NDEBUG)
113 # define CLANG_DISABLE_OPTIMIZATION_ATTRIBUTE __attribute__((optnone))
115 # define CLANG_DISABLE_OPTIMIZATION_ATTRIBUTE
119 #endif /* CUDA_ARCH_UTILS_CUH_ */