1/*===---- cmath - CUDA wrapper for <cmath> ---------------------------------=== 2 * 3 * Permission is hereby granted, free of charge, to any person obtaining a copy 4 * of this software and associated documentation files (the "Software"), to deal 5 * in the Software without restriction, including without limitation the rights 6 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 7 * copies of the Software, and to permit persons to whom the Software is 8 * furnished to do so, subject to the following conditions: 9 * 10 * The above copyright notice and this permission notice shall be included in 11 * all copies or substantial portions of the Software. 12 * 13 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 14 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 15 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 16 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 18 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 19 * THE SOFTWARE. 20 * 21 *===-----------------------------------------------------------------------=== 22 */ 23 24#ifndef __CLANG_CUDA_WRAPPERS_CMATH 25#define __CLANG_CUDA_WRAPPERS_CMATH 26 27#if __has_include_next(<cmath>) 28#include_next <cmath> 29#else 30#error "Could not find standard C++ header 'cmath'. Add -v to your compilation command to check the include paths being searched. You may need to install the appropriate standard C++ library package corresponding to the search path." 31#endif 32 33#if defined(_LIBCPP_STD_VER) 34 35// libc++ will need long double variants of these functions, but CUDA does not 36// provide them. We'll provide their declarations, which should allow the 37// headers to parse, but would not allow accidental use of them on a GPU. 38 39__attribute__((device)) long double logb(long double); 40__attribute__((device)) long double scalbn(long double, int); 41 42namespace std { 43 44// For __constexpr_fmin/fmax we only need device-side overloads before c++14 45// where they are not constexpr. 46#if _LIBCPP_STD_VER < 14 47 48__attribute__((device)) 49inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 float __constexpr_fmax(float __x, float __y) _NOEXCEPT { 50 return __builtin_fmaxf(__x, __y); 51} 52 53__attribute__((device)) 54inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 double __constexpr_fmax(double __x, double __y) _NOEXCEPT { 55 return __builtin_fmax(__x, __y); 56} 57 58__attribute__((device)) 59inline _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 long double 60__constexpr_fmax(long double __x, long double __y) _NOEXCEPT { 61 return __builtin_fmaxl(__x, __y); 62} 63 64template <class _Tp, class _Up, __enable_if_t<is_arithmetic<_Tp>::value && is_arithmetic<_Up>::value, int> = 0> 65__attribute__((device)) 66_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 typename __promote<_Tp, _Up>::type 67__constexpr_fmax(_Tp __x, _Up __y) _NOEXCEPT { 68 using __result_type = typename __promote<_Tp, _Up>::type; 69 return std::__constexpr_fmax(static_cast<__result_type>(__x), static_cast<__result_type>(__y)); 70} 71#endif // _LIBCPP_STD_VER < 14 72 73// For logb/scalbn templates we must always provide device overloads because 74// libc++ implementation uses __builtin_XXX which gets translated into a libcall 75// which we can't handle on GPU. We need to forward those to CUDA-provided 76// implementations. 77 78template <class _Tp> 79__attribute__((device)) 80_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 _Tp __constexpr_logb(_Tp __x) { 81 return ::logb(__x); 82} 83 84template <class _Tp> 85__attribute__((device)) 86_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 _Tp __constexpr_scalbn(_Tp __x, int __exp) { 87 return ::scalbn(__x, __exp); 88} 89 90} // namespace std// 91 92#endif // _LIBCPP_STD_VER 93 94#endif // include guard 95