| 1 | //===-- Implementation of the C++20 bit header -----------------*- 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 | // This is inspired by LLVM ADT/bit.h header. |
| 9 | // Some functions are missing, we can add them as needed (popcount, byteswap). |
| 10 | |
| 11 | #ifndef LLVM_LIBC_SRC___SUPPORT_CPP_BIT_H |
| 12 | #define LLVM_LIBC_SRC___SUPPORT_CPP_BIT_H |
| 13 | |
| 14 | #include "src/__support/CPP/limits.h" // numeric_limits |
| 15 | #include "src/__support/CPP/type_traits.h" |
| 16 | #include "src/__support/macros/attributes.h" |
| 17 | #include "src/__support/macros/config.h" |
| 18 | #include "src/__support/macros/sanitizer.h" |
| 19 | |
| 20 | #include <stdint.h> |
| 21 | |
| 22 | namespace LIBC_NAMESPACE_DECL { |
| 23 | namespace cpp { |
| 24 | |
| 25 | #if __has_builtin(__builtin_memcpy_inline) |
| 26 | #define LLVM_LIBC_HAS_BUILTIN_MEMCPY_INLINE |
| 27 | #endif |
| 28 | |
| 29 | // This implementation of bit_cast requires trivially-constructible To, to avoid |
| 30 | // UB in the implementation. |
| 31 | template <typename To, typename From> |
| 32 | LIBC_INLINE constexpr cpp::enable_if_t< |
| 33 | (sizeof(To) == sizeof(From)) && |
| 34 | cpp::is_trivially_constructible<To>::value && |
| 35 | cpp::is_trivially_copyable<To>::value && |
| 36 | cpp::is_trivially_copyable<From>::value, |
| 37 | To> |
| 38 | bit_cast(const From &from) { |
| 39 | MSAN_UNPOISON(&from, sizeof(From)); |
| 40 | #if __has_builtin(__builtin_bit_cast) |
| 41 | return __builtin_bit_cast(To, from); |
| 42 | #else |
| 43 | To to; |
| 44 | char *dst = reinterpret_cast<char *>(&to); |
| 45 | const char *src = reinterpret_cast<const char *>(&from); |
| 46 | #if __has_builtin(__builtin_memcpy_inline) |
| 47 | __builtin_memcpy_inline(dst, src, sizeof(To)); |
| 48 | #else |
| 49 | for (unsigned i = 0; i < sizeof(To); ++i) |
| 50 | dst[i] = src[i]; |
| 51 | #endif // __has_builtin(__builtin_memcpy_inline) |
| 52 | return to; |
| 53 | #endif // __has_builtin(__builtin_bit_cast) |
| 54 | } |
| 55 | |
| 56 | template <typename T> |
| 57 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, |
| 58 | bool> |
| 59 | has_single_bit(T value) { |
| 60 | return (value != 0) && ((value & (value - 1)) == 0); |
| 61 | } |
| 62 | |
| 63 | // A temporary macro to add template function specialization when compiler |
| 64 | // builtin is available. |
| 65 | #define ADD_SPECIALIZATION(NAME, TYPE, BUILTIN) \ |
| 66 | template <> [[nodiscard]] LIBC_INLINE constexpr int NAME<TYPE>(TYPE value) { \ |
| 67 | static_assert(cpp::is_unsigned_v<TYPE>); \ |
| 68 | return value == 0 ? cpp::numeric_limits<TYPE>::digits : BUILTIN(value); \ |
| 69 | } |
| 70 | |
| 71 | /// Count number of 0's from the least significant bit to the most |
| 72 | /// stopping at the first 1. |
| 73 | /// |
| 74 | /// Only unsigned integral types are allowed. |
| 75 | /// |
| 76 | /// Returns cpp::numeric_limits<T>::digits on an input of 0. |
| 77 | // clang-19+, gcc-14+ |
| 78 | #if __has_builtin(__builtin_ctzg) |
| 79 | template <typename T> |
| 80 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 81 | countr_zero(T value) { |
| 82 | return __builtin_ctzg(value, cpp::numeric_limits<T>::digits); |
| 83 | } |
| 84 | #else |
| 85 | template <typename T> |
| 86 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 87 | countr_zero(T value) { |
| 88 | if (!value) |
| 89 | return cpp::numeric_limits<T>::digits; |
| 90 | if (value & 0x1) |
| 91 | return 0; |
| 92 | // Bisection method. |
| 93 | unsigned zero_bits = 0; |
| 94 | unsigned shift = cpp::numeric_limits<T>::digits >> 1; |
| 95 | T mask = cpp::numeric_limits<T>::max() >> shift; |
| 96 | while (shift) { |
| 97 | if ((value & mask) == 0) { |
| 98 | value >>= shift; |
| 99 | zero_bits |= shift; |
| 100 | } |
| 101 | shift >>= 1; |
| 102 | mask >>= shift; |
| 103 | } |
| 104 | return static_cast<int>(zero_bits); |
| 105 | } |
| 106 | #if __has_builtin(__builtin_ctzs) |
| 107 | ADD_SPECIALIZATION(countr_zero, unsigned short, __builtin_ctzs) |
| 108 | #endif |
| 109 | ADD_SPECIALIZATION(countr_zero, unsigned int, __builtin_ctz) |
| 110 | ADD_SPECIALIZATION(countr_zero, unsigned long, __builtin_ctzl) |
| 111 | ADD_SPECIALIZATION(countr_zero, unsigned long long, __builtin_ctzll) |
| 112 | #endif // __has_builtin(__builtin_ctzg) |
| 113 | |
| 114 | /// Count number of 0's from the most significant bit to the least |
| 115 | /// stopping at the first 1. |
| 116 | /// |
| 117 | /// Only unsigned integral types are allowed. |
| 118 | /// |
| 119 | /// Returns cpp::numeric_limits<T>::digits on an input of 0. |
| 120 | // clang-19+, gcc-14+ |
| 121 | #if __has_builtin(__builtin_clzg) |
| 122 | template <typename T> |
| 123 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 124 | countl_zero(T value) { |
| 125 | return __builtin_clzg(value, cpp::numeric_limits<T>::digits); |
| 126 | } |
| 127 | #else |
| 128 | template <typename T> |
| 129 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 130 | countl_zero(T value) { |
| 131 | if (!value) |
| 132 | return cpp::numeric_limits<T>::digits; |
| 133 | // Bisection method. |
| 134 | unsigned zero_bits = 0; |
| 135 | for (unsigned shift = cpp::numeric_limits<T>::digits >> 1; shift; |
| 136 | shift >>= 1) { |
| 137 | T tmp = value >> shift; |
| 138 | if (tmp) |
| 139 | value = tmp; |
| 140 | else |
| 141 | zero_bits |= shift; |
| 142 | } |
| 143 | return static_cast<int>(zero_bits); |
| 144 | } |
| 145 | #if __has_builtin(__builtin_clzs) |
| 146 | ADD_SPECIALIZATION(countl_zero, unsigned short, __builtin_clzs) |
| 147 | #endif |
| 148 | ADD_SPECIALIZATION(countl_zero, unsigned int, __builtin_clz) |
| 149 | ADD_SPECIALIZATION(countl_zero, unsigned long, __builtin_clzl) |
| 150 | ADD_SPECIALIZATION(countl_zero, unsigned long long, __builtin_clzll) |
| 151 | #endif // __has_builtin(__builtin_clzg) |
| 152 | |
| 153 | #undef ADD_SPECIALIZATION |
| 154 | |
| 155 | /// Count the number of ones from the most significant bit to the first |
| 156 | /// zero bit. |
| 157 | /// |
| 158 | /// Ex. countl_one(0xFF0FFF00) == 8. |
| 159 | /// Only unsigned integral types are allowed. |
| 160 | /// |
| 161 | /// Returns cpp::numeric_limits<T>::digits on an input of all ones. |
| 162 | template <typename T> |
| 163 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 164 | countl_one(T value) { |
| 165 | return cpp::countl_zero<T>(static_cast<T>(~value)); |
| 166 | } |
| 167 | |
| 168 | /// Count the number of ones from the least significant bit to the first |
| 169 | /// zero bit. |
| 170 | /// |
| 171 | /// Ex. countr_one(0x00FF00FF) == 8. |
| 172 | /// Only unsigned integral types are allowed. |
| 173 | /// |
| 174 | /// Returns cpp::numeric_limits<T>::digits on an input of all ones. |
| 175 | template <typename T> |
| 176 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 177 | countr_one(T value) { |
| 178 | return cpp::countr_zero<T>(static_cast<T>(~value)); |
| 179 | } |
| 180 | |
| 181 | /// Returns the number of bits needed to represent value if value is nonzero. |
| 182 | /// Returns 0 otherwise. |
| 183 | /// |
| 184 | /// Ex. bit_width(5) == 3. |
| 185 | template <typename T> |
| 186 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 187 | bit_width(T value) { |
| 188 | return cpp::numeric_limits<T>::digits - cpp::countl_zero(value); |
| 189 | } |
| 190 | |
| 191 | /// Returns the largest integral power of two no greater than value if value is |
| 192 | /// nonzero. Returns 0 otherwise. |
| 193 | /// |
| 194 | /// Ex. bit_floor(5) == 4. |
| 195 | template <typename T> |
| 196 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, T> |
| 197 | bit_floor(T value) { |
| 198 | if (!value) |
| 199 | return 0; |
| 200 | return static_cast<T>(T(1) << (cpp::bit_width(value) - 1)); |
| 201 | } |
| 202 | |
| 203 | /// Returns the smallest integral power of two no smaller than value if value is |
| 204 | /// nonzero. Returns 1 otherwise. |
| 205 | /// |
| 206 | /// Ex. bit_ceil(5) == 8. |
| 207 | /// |
| 208 | /// The return value is undefined if the input is larger than the largest power |
| 209 | /// of two representable in T. |
| 210 | template <typename T> |
| 211 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, T> |
| 212 | bit_ceil(T value) { |
| 213 | if (value < 2) |
| 214 | return 1; |
| 215 | return static_cast<T>(T(1) << cpp::bit_width(value - 1U)); |
| 216 | } |
| 217 | |
| 218 | // Rotate algorithms make use of "Safe, Efficient, and Portable Rotate in C/C++" |
| 219 | // from https://blog.regehr.org/archives/1063. |
| 220 | |
| 221 | // Forward-declare rotr so that rotl can use it. |
| 222 | template <typename T> |
| 223 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, T> |
| 224 | rotr(T value, int rotate); |
| 225 | |
| 226 | template <typename T> |
| 227 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, T> |
| 228 | rotl(T value, int rotate) { |
| 229 | constexpr int N = cpp::numeric_limits<T>::digits; |
| 230 | rotate = rotate % N; |
| 231 | if (!rotate) |
| 232 | return value; |
| 233 | if (rotate < 0) |
| 234 | return cpp::rotr<T>(value, -rotate); |
| 235 | return static_cast<T>((value << rotate) | (value >> (N - rotate))); |
| 236 | } |
| 237 | |
| 238 | template <typename T> |
| 239 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, T> |
| 240 | rotr(T value, int rotate) { |
| 241 | constexpr int N = cpp::numeric_limits<T>::digits; |
| 242 | rotate = rotate % N; |
| 243 | if (!rotate) |
| 244 | return value; |
| 245 | if (rotate < 0) |
| 246 | return cpp::rotl<T>(value, -rotate); |
| 247 | return static_cast<T>((value >> rotate) | (value << (N - rotate))); |
| 248 | } |
| 249 | |
| 250 | // TODO: Do we need this function at all? How is it different from |
| 251 | // 'static_cast'? |
| 252 | template <class To, class From> |
| 253 | LIBC_INLINE constexpr To bit_or_static_cast(const From &from) { |
| 254 | if constexpr (sizeof(To) == sizeof(From)) { |
| 255 | return bit_cast<To>(from); |
| 256 | } else { |
| 257 | return static_cast<To>(from); |
| 258 | } |
| 259 | } |
| 260 | |
| 261 | /// Count number of 1's aka population count or Hamming weight. |
| 262 | /// |
| 263 | /// Only unsigned integral types are allowed. |
| 264 | // clang-19+, gcc-14+ |
| 265 | #if __has_builtin(__builtin_popcountg) |
| 266 | template <typename T> |
| 267 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 268 | popcount(T value) { |
| 269 | return __builtin_popcountg(value); |
| 270 | } |
| 271 | #else // !__has_builtin(__builtin_popcountg) |
| 272 | template <typename T> |
| 273 | [[nodiscard]] LIBC_INLINE constexpr cpp::enable_if_t<cpp::is_unsigned_v<T>, int> |
| 274 | popcount(T value) { |
| 275 | int count = 0; |
| 276 | while (value) { |
| 277 | value &= value - 1; |
| 278 | ++count; |
| 279 | } |
| 280 | return count; |
| 281 | } |
| 282 | #define ADD_SPECIALIZATION(TYPE, BUILTIN) \ |
| 283 | template <> \ |
| 284 | [[nodiscard]] LIBC_INLINE constexpr int popcount<TYPE>(TYPE value) { \ |
| 285 | return BUILTIN(value); \ |
| 286 | } |
| 287 | ADD_SPECIALIZATION(unsigned char, __builtin_popcount) |
| 288 | ADD_SPECIALIZATION(unsigned short, __builtin_popcount) |
| 289 | ADD_SPECIALIZATION(unsigned, __builtin_popcount) |
| 290 | ADD_SPECIALIZATION(unsigned long, __builtin_popcountl) |
| 291 | ADD_SPECIALIZATION(unsigned long long, __builtin_popcountll) |
| 292 | #endif // __builtin_popcountg |
| 293 | #undef ADD_SPECIALIZATION |
| 294 | |
| 295 | } // namespace cpp |
| 296 | } // namespace LIBC_NAMESPACE_DECL |
| 297 | |
| 298 | #endif // LLVM_LIBC_SRC___SUPPORT_CPP_BIT_H |
| 299 | |