| 1 |
/*===-- __clang_openmp_device_functions.h - OpenMP math declares ------ c++ -=== |
| 2 |
* |
| 3 |
* Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 |
* See https://llvm.org/LICENSE.txt for license information. |
| 5 |
* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 |
* |
| 7 |
*===-----------------------------------------------------------------------=== |
| 8 |
*/ |
| 9 |
|
| 10 |
#ifndef __CLANG_OPENMP_CMATH_H__ |
| 11 |
#define __CLANG_OPENMP_CMATH_H__ |
| 12 |
|
| 13 |
#ifndef _OPENMP |
| 14 |
#error "This file is for OpenMP compilation only." |
| 15 |
#endif |
| 16 |
|
| 17 |
#include_next <cmath> |
| 18 |
|
| 19 |
// Make sure we include our math.h overlay, it probably happend already but we |
| 20 |
// need to be sure. |
| 21 |
#include <math.h> |
| 22 |
|
| 23 |
// We (might) need cstdlib because __clang_cuda_cmath.h below declares `abs` |
| 24 |
// which might live in cstdlib. |
| 25 |
#include <cstdlib> |
| 26 |
|
| 27 |
// We need limits because __clang_cuda_cmath.h below uses `std::numeric_limit`. |
| 28 |
#include <limits> |
| 29 |
|
| 30 |
#pragma omp begin declare variant match( \ |
| 31 |
device = {arch(nvptx, nvptx64)}, implementation = {extension(match_any, allow_templates)}) |
| 32 |
|
| 33 |
#define __CUDA__ |
| 34 |
#define __OPENMP_NVPTX__ |
| 35 |
#include <__clang_cuda_cmath.h> |
| 36 |
#undef __OPENMP_NVPTX__ |
| 37 |
#undef __CUDA__ |
| 38 |
|
| 39 |
// Overloads not provided by the CUDA wrappers but by the CUDA system headers. |
| 40 |
// Since we do not include the latter we define them ourselves. |
| 41 |
#define __DEVICE__ static constexpr __attribute__((always_inline, nothrow)) |
| 42 |
|
| 43 |
__DEVICE__ float acosh(float __x) { return ::acoshf(__x); } |
| 44 |
__DEVICE__ float asinh(float __x) { return ::asinhf(__x); } |
| 45 |
__DEVICE__ float atanh(float __x) { return ::atanhf(__x); } |
| 46 |
__DEVICE__ float cbrt(float __x) { return ::cbrtf(__x); } |
| 47 |
__DEVICE__ float erf(float __x) { return ::erff(__x); } |
| 48 |
__DEVICE__ float erfc(float __x) { return ::erfcf(__x); } |
| 49 |
__DEVICE__ float exp2(float __x) { return ::exp2f(__x); } |
| 50 |
__DEVICE__ float expm1(float __x) { return ::expm1f(__x); } |
| 51 |
__DEVICE__ float fdim(float __x, float __y) { return ::fdimf(__x, __y); } |
| 52 |
__DEVICE__ float hypot(float __x, float __y) { return ::hypotf(__x, __y); } |
| 53 |
__DEVICE__ int ilogb(float __x) { return ::ilogbf(__x); } |
| 54 |
__DEVICE__ float lgamma(float __x) { return ::lgammaf(__x); } |
| 55 |
__DEVICE__ long long int llrint(float __x) { return ::llrintf(__x); } |
| 56 |
__DEVICE__ long long int llround(float __x) { return ::llroundf(__x); } |
| 57 |
__DEVICE__ float log1p(float __x) { return ::log1pf(__x); } |
| 58 |
__DEVICE__ float log2(float __x) { return ::log2f(__x); } |
| 59 |
__DEVICE__ float logb(float __x) { return ::logbf(__x); } |
| 60 |
__DEVICE__ long int lrint(float __x) { return ::lrintf(__x); } |
| 61 |
__DEVICE__ long int lround(float __x) { return ::lroundf(__x); } |
| 62 |
__DEVICE__ float nextafter(float __x, float __y) { |
| 63 |
return ::nextafterf(__x, __y); |
| 64 |
} |
| 65 |
__DEVICE__ float remainder(float __x, float __y) { |
| 66 |
return ::remainderf(__x, __y); |
| 67 |
} |
| 68 |
__DEVICE__ float scalbln(float __x, long int __y) { |
| 69 |
return ::scalblnf(__x, __y); |
| 70 |
} |
| 71 |
__DEVICE__ float scalbn(float __x, int __y) { return ::scalbnf(__x, __y); } |
| 72 |
__DEVICE__ float tgamma(float __x) { return ::tgammaf(__x); } |
| 73 |
|
| 74 |
#undef __DEVICE__ |
| 75 |
|
| 76 |
#pragma omp end declare variant |
| 77 |
|
| 78 |
#ifdef __AMDGCN__ |
| 79 |
#pragma omp begin declare variant match(device = {arch(amdgcn)}) |
| 80 |
|
| 81 |
#pragma push_macro("__constant__") |
| 82 |
#define __constant__ __attribute__((constant)) |
| 83 |
#define __OPENMP_AMDGCN__ |
| 84 |
|
| 85 |
#include <__clang_hip_cmath.h> |
| 86 |
|
| 87 |
#pragma pop_macro("__constant__") |
| 88 |
#undef __OPENMP_AMDGCN__ |
| 89 |
|
| 90 |
// Define overloads otherwise which are absent |
| 91 |
#define __DEVICE__ static constexpr __attribute__((always_inline, nothrow)) |
| 92 |
|
| 93 |
__DEVICE__ float acos(float __x) { return ::acosf(__x); } |
| 94 |
__DEVICE__ float acosh(float __x) { return ::acoshf(__x); } |
| 95 |
__DEVICE__ float asin(float __x) { return ::asinf(__x); } |
| 96 |
__DEVICE__ float asinh(float __x) { return ::asinhf(__x); } |
| 97 |
__DEVICE__ float atan(float __x) { return ::atanf(__x); } |
| 98 |
__DEVICE__ float atan2(float __x, float __y) { return ::atan2f(__x, __y); } |
| 99 |
__DEVICE__ float atanh(float __x) { return ::atanhf(__x); } |
| 100 |
__DEVICE__ float cbrt(float __x) { return ::cbrtf(__x); } |
| 101 |
__DEVICE__ float cosh(float __x) { return ::coshf(__x); } |
| 102 |
__DEVICE__ float erf(float __x) { return ::erff(__x); } |
| 103 |
__DEVICE__ float erfc(float __x) { return ::erfcf(__x); } |
| 104 |
__DEVICE__ float exp2(float __x) { return ::exp2f(__x); } |
| 105 |
__DEVICE__ float expm1(float __x) { return ::expm1f(__x); } |
| 106 |
__DEVICE__ float fdim(float __x, float __y) { return ::fdimf(__x, __y); } |
| 107 |
__DEVICE__ float hypot(float __x, float __y) { return ::hypotf(__x, __y); } |
| 108 |
__DEVICE__ int ilogb(float __x) { return ::ilogbf(__x); } |
| 109 |
__DEVICE__ float ldexp(float __arg, int __exp) { |
| 110 |
return ::ldexpf(__arg, __exp); |
| 111 |
} |
| 112 |
__DEVICE__ float lgamma(float __x) { return ::lgammaf(__x); } |
| 113 |
__DEVICE__ float log1p(float __x) { return ::log1pf(__x); } |
| 114 |
__DEVICE__ float logb(float __x) { return ::logbf(__x); } |
| 115 |
__DEVICE__ float nextafter(float __x, float __y) { |
| 116 |
return ::nextafterf(__x, __y); |
| 117 |
} |
| 118 |
__DEVICE__ float remainder(float __x, float __y) { |
| 119 |
return ::remainderf(__x, __y); |
| 120 |
} |
| 121 |
__DEVICE__ float scalbn(float __x, int __y) { return ::scalbnf(__x, __y); } |
| 122 |
__DEVICE__ float sinh(float __x) { return ::sinhf(__x); } |
| 123 |
__DEVICE__ float tan(float __x) { return ::tanf(__x); } |
| 124 |
__DEVICE__ float tanh(float __x) { return ::tanhf(__x); } |
| 125 |
__DEVICE__ float tgamma(float __x) { return ::tgammaf(__x); } |
| 126 |
|
| 127 |
#undef __DEVICE__ |
| 128 |
|
| 129 |
#pragma omp end declare variant |
| 130 |
#endif // __AMDGCN__ |
| 131 |
|
| 132 |
#endif |