Spaces:
Running
Running
File size: 8,272 Bytes
c61ccee |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 |
#pragma once
#include <c10/util/BFloat16.h>
#include <c10/util/Half.h>
C10_CLANG_DIAGNOSTIC_PUSH()
#if C10_CLANG_HAS_WARNING("-Wimplicit-float-conversion")
C10_CLANG_DIAGNOSTIC_IGNORE("-Wimplicit-float-conversion")
#endif
namespace std {
template <typename T>
struct is_reduced_floating_point
: std::integral_constant<
bool,
std::is_same_v<T, c10::Half> || std::is_same_v<T, c10::BFloat16>> {};
template <typename T>
constexpr bool is_reduced_floating_point_v =
is_reduced_floating_point<T>::value;
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T acos(T a) {
return std::acos(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T asin(T a) {
return std::asin(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T atan(T a) {
return std::atan(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T atanh(T a) {
return std::atanh(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T erf(T a) {
return std::erf(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T erfc(T a) {
return std::erfc(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T exp(T a) {
return std::exp(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T expm1(T a) {
return std::expm1(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T log(T a) {
return std::log(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T log10(T a) {
return std::log10(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T log1p(T a) {
return std::log1p(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T log2(T a) {
return std::log2(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T ceil(T a) {
return std::ceil(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T cos(T a) {
return std::cos(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T floor(T a) {
return std::floor(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T nearbyint(T a) {
return std::nearbyint(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T sin(T a) {
return std::sin(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T tan(T a) {
return std::tan(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T sinh(T a) {
return std::sinh(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T cosh(T a) {
return std::cosh(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T tanh(T a) {
return std::tanh(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T trunc(T a) {
return std::trunc(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T lgamma(T a) {
return std::lgamma(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T sqrt(T a) {
return std::sqrt(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T rsqrt(T a) {
return 1.0 / std::sqrt(float(a));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T abs(T a) {
return std::abs(float(a));
}
#if defined(_MSC_VER) && defined(__CUDACC__)
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T pow(T a, double b) {
return std::pow(float(a), float(b));
}
#else
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T pow(T a, double b) {
return std::pow(float(a), b);
}
#endif
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T pow(T a, T b) {
return std::pow(float(a), float(b));
}
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
inline T fmod(T a, T b) {
return std::fmod(float(a), float(b));
}
/*
The following function is inspired from the implementation in `musl`
Link to License: https://git.musl-libc.org/cgit/musl/tree/COPYRIGHT
----------------------------------------------------------------------
Copyright © 2005-2020 Rich Felker, et al.
Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
"Software"), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:
The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
----------------------------------------------------------------------
*/
template <
typename T,
typename std::enable_if_t<is_reduced_floating_point_v<T>, int> = 0>
C10_HOST_DEVICE inline T nextafter(T from, T to) {
// Reference:
// https://git.musl-libc.org/cgit/musl/tree/src/math/nextafter.c
using int_repr_t = uint16_t;
using float_t = T;
constexpr uint8_t bits = 16;
union {
float_t f;
int_repr_t i;
} ufrom = {from}, uto = {to};
// get a mask to get the sign bit i.e. MSB
int_repr_t sign_mask = int_repr_t{1} << (bits - 1);
// short-circuit: if either is NaN, return NaN
if (from != from || to != to) {
return from + to;
}
// short-circuit: if they are exactly the same.
if (ufrom.i == uto.i) {
return from;
}
// mask the sign-bit to zero i.e. positive
// equivalent to abs(x)
int_repr_t abs_from = ufrom.i & ~sign_mask;
int_repr_t abs_to = uto.i & ~sign_mask;
if (abs_from == 0) {
// if both are zero but with different sign,
// preserve the sign of `to`.
if (abs_to == 0) {
return to;
}
// smallest subnormal with sign of `to`.
ufrom.i = (uto.i & sign_mask) | int_repr_t{1};
return ufrom.f;
}
// if abs(from) > abs(to) or sign(from) != sign(to)
if (abs_from > abs_to || ((ufrom.i ^ uto.i) & sign_mask)) {
ufrom.i--;
} else {
ufrom.i++;
}
return ufrom.f;
}
} // namespace std
C10_CLANG_DIAGNOSTIC_POP()
|