Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
| 2 | #ifndef _LINUX_MATH64_H |
| 3 | #define _LINUX_MATH64_H |
| 4 | |
| 5 | #include <linux/types.h> |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 6 | #include <vdso/math64.h> |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 7 | #include <asm/div64.h> |
| 8 | |
| 9 | #if BITS_PER_LONG == 64 |
| 10 | |
| 11 | #define div64_long(x, y) div64_s64((x), (y)) |
| 12 | #define div64_ul(x, y) div64_u64((x), (y)) |
| 13 | |
| 14 | /** |
| 15 | * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder |
| 16 | * @dividend: unsigned 64bit dividend |
| 17 | * @divisor: unsigned 32bit divisor |
| 18 | * @remainder: pointer to unsigned 32bit remainder |
| 19 | * |
| 20 | * Return: sets ``*remainder``, then returns dividend / divisor |
| 21 | * |
| 22 | * This is commonly provided by 32bit archs to provide an optimized 64bit |
| 23 | * divide. |
| 24 | */ |
| 25 | static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder) |
| 26 | { |
| 27 | *remainder = dividend % divisor; |
| 28 | return dividend / divisor; |
| 29 | } |
| 30 | |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 31 | /* |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 32 | * div_s64_rem - signed 64bit divide with 32bit divisor with remainder |
| 33 | * @dividend: signed 64bit dividend |
| 34 | * @divisor: signed 32bit divisor |
| 35 | * @remainder: pointer to signed 32bit remainder |
| 36 | * |
| 37 | * Return: sets ``*remainder``, then returns dividend / divisor |
| 38 | */ |
| 39 | static inline s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder) |
| 40 | { |
| 41 | *remainder = dividend % divisor; |
| 42 | return dividend / divisor; |
| 43 | } |
| 44 | |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 45 | /* |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 46 | * div64_u64_rem - unsigned 64bit divide with 64bit divisor and remainder |
| 47 | * @dividend: unsigned 64bit dividend |
| 48 | * @divisor: unsigned 64bit divisor |
| 49 | * @remainder: pointer to unsigned 64bit remainder |
| 50 | * |
| 51 | * Return: sets ``*remainder``, then returns dividend / divisor |
| 52 | */ |
| 53 | static inline u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder) |
| 54 | { |
| 55 | *remainder = dividend % divisor; |
| 56 | return dividend / divisor; |
| 57 | } |
| 58 | |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 59 | /* |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 60 | * div64_u64 - unsigned 64bit divide with 64bit divisor |
| 61 | * @dividend: unsigned 64bit dividend |
| 62 | * @divisor: unsigned 64bit divisor |
| 63 | * |
| 64 | * Return: dividend / divisor |
| 65 | */ |
| 66 | static inline u64 div64_u64(u64 dividend, u64 divisor) |
| 67 | { |
| 68 | return dividend / divisor; |
| 69 | } |
| 70 | |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 71 | /* |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 72 | * div64_s64 - signed 64bit divide with 64bit divisor |
| 73 | * @dividend: signed 64bit dividend |
| 74 | * @divisor: signed 64bit divisor |
| 75 | * |
| 76 | * Return: dividend / divisor |
| 77 | */ |
| 78 | static inline s64 div64_s64(s64 dividend, s64 divisor) |
| 79 | { |
| 80 | return dividend / divisor; |
| 81 | } |
| 82 | |
| 83 | #elif BITS_PER_LONG == 32 |
| 84 | |
| 85 | #define div64_long(x, y) div_s64((x), (y)) |
| 86 | #define div64_ul(x, y) div_u64((x), (y)) |
| 87 | |
| 88 | #ifndef div_u64_rem |
| 89 | static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder) |
| 90 | { |
| 91 | *remainder = do_div(dividend, divisor); |
| 92 | return dividend; |
| 93 | } |
| 94 | #endif |
| 95 | |
| 96 | #ifndef div_s64_rem |
| 97 | extern s64 div_s64_rem(s64 dividend, s32 divisor, s32 *remainder); |
| 98 | #endif |
| 99 | |
| 100 | #ifndef div64_u64_rem |
| 101 | extern u64 div64_u64_rem(u64 dividend, u64 divisor, u64 *remainder); |
| 102 | #endif |
| 103 | |
| 104 | #ifndef div64_u64 |
| 105 | extern u64 div64_u64(u64 dividend, u64 divisor); |
| 106 | #endif |
| 107 | |
| 108 | #ifndef div64_s64 |
| 109 | extern s64 div64_s64(s64 dividend, s64 divisor); |
| 110 | #endif |
| 111 | |
| 112 | #endif /* BITS_PER_LONG */ |
| 113 | |
| 114 | /** |
| 115 | * div_u64 - unsigned 64bit divide with 32bit divisor |
| 116 | * @dividend: unsigned 64bit dividend |
| 117 | * @divisor: unsigned 32bit divisor |
| 118 | * |
| 119 | * This is the most common 64bit divide and should be used if possible, |
| 120 | * as many 32bit archs can optimize this variant better than a full 64bit |
| 121 | * divide. |
| 122 | */ |
| 123 | #ifndef div_u64 |
| 124 | static inline u64 div_u64(u64 dividend, u32 divisor) |
| 125 | { |
| 126 | u32 remainder; |
| 127 | return div_u64_rem(dividend, divisor, &remainder); |
| 128 | } |
| 129 | #endif |
| 130 | |
| 131 | /** |
| 132 | * div_s64 - signed 64bit divide with 32bit divisor |
| 133 | * @dividend: signed 64bit dividend |
| 134 | * @divisor: signed 32bit divisor |
| 135 | */ |
| 136 | #ifndef div_s64 |
| 137 | static inline s64 div_s64(s64 dividend, s32 divisor) |
| 138 | { |
| 139 | s32 remainder; |
| 140 | return div_s64_rem(dividend, divisor, &remainder); |
| 141 | } |
| 142 | #endif |
| 143 | |
| 144 | u32 iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder); |
| 145 | |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 146 | #ifndef mul_u32_u32 |
| 147 | /* |
| 148 | * Many a GCC version messes this up and generates a 64x64 mult :-( |
| 149 | */ |
| 150 | static inline u64 mul_u32_u32(u32 a, u32 b) |
| 151 | { |
| 152 | return (u64)a * b; |
| 153 | } |
| 154 | #endif |
| 155 | |
| 156 | #if defined(CONFIG_ARCH_SUPPORTS_INT128) && defined(__SIZEOF_INT128__) |
| 157 | |
| 158 | #ifndef mul_u64_u32_shr |
| 159 | static inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift) |
| 160 | { |
| 161 | return (u64)(((unsigned __int128)a * mul) >> shift); |
| 162 | } |
| 163 | #endif /* mul_u64_u32_shr */ |
| 164 | |
| 165 | #ifndef mul_u64_u64_shr |
| 166 | static inline u64 mul_u64_u64_shr(u64 a, u64 mul, unsigned int shift) |
| 167 | { |
| 168 | return (u64)(((unsigned __int128)a * mul) >> shift); |
| 169 | } |
| 170 | #endif /* mul_u64_u64_shr */ |
| 171 | |
| 172 | #else |
| 173 | |
| 174 | #ifndef mul_u64_u32_shr |
| 175 | static inline u64 mul_u64_u32_shr(u64 a, u32 mul, unsigned int shift) |
| 176 | { |
| 177 | u32 ah, al; |
| 178 | u64 ret; |
| 179 | |
| 180 | al = a; |
| 181 | ah = a >> 32; |
| 182 | |
| 183 | ret = mul_u32_u32(al, mul) >> shift; |
| 184 | if (ah) |
| 185 | ret += mul_u32_u32(ah, mul) << (32 - shift); |
| 186 | |
| 187 | return ret; |
| 188 | } |
| 189 | #endif /* mul_u64_u32_shr */ |
| 190 | |
| 191 | #ifndef mul_u64_u64_shr |
| 192 | static inline u64 mul_u64_u64_shr(u64 a, u64 b, unsigned int shift) |
| 193 | { |
| 194 | union { |
| 195 | u64 ll; |
| 196 | struct { |
| 197 | #ifdef __BIG_ENDIAN |
| 198 | u32 high, low; |
| 199 | #else |
| 200 | u32 low, high; |
| 201 | #endif |
| 202 | } l; |
| 203 | } rl, rm, rn, rh, a0, b0; |
| 204 | u64 c; |
| 205 | |
| 206 | a0.ll = a; |
| 207 | b0.ll = b; |
| 208 | |
| 209 | rl.ll = mul_u32_u32(a0.l.low, b0.l.low); |
| 210 | rm.ll = mul_u32_u32(a0.l.low, b0.l.high); |
| 211 | rn.ll = mul_u32_u32(a0.l.high, b0.l.low); |
| 212 | rh.ll = mul_u32_u32(a0.l.high, b0.l.high); |
| 213 | |
| 214 | /* |
| 215 | * Each of these lines computes a 64-bit intermediate result into "c", |
| 216 | * starting at bits 32-95. The low 32-bits go into the result of the |
| 217 | * multiplication, the high 32-bits are carried into the next step. |
| 218 | */ |
| 219 | rl.l.high = c = (u64)rl.l.high + rm.l.low + rn.l.low; |
| 220 | rh.l.low = c = (c >> 32) + rm.l.high + rn.l.high + rh.l.low; |
| 221 | rh.l.high = (c >> 32) + rh.l.high; |
| 222 | |
| 223 | /* |
| 224 | * The 128-bit result of the multiplication is in rl.ll and rh.ll, |
| 225 | * shift it right and throw away the high part of the result. |
| 226 | */ |
| 227 | if (shift == 0) |
| 228 | return rl.ll; |
| 229 | if (shift < 64) |
| 230 | return (rl.ll >> shift) | (rh.ll << (64 - shift)); |
| 231 | return rh.ll >> (shift & 63); |
| 232 | } |
| 233 | #endif /* mul_u64_u64_shr */ |
| 234 | |
| 235 | #endif |
| 236 | |
| 237 | #ifndef mul_u64_u32_div |
| 238 | static inline u64 mul_u64_u32_div(u64 a, u32 mul, u32 divisor) |
| 239 | { |
| 240 | union { |
| 241 | u64 ll; |
| 242 | struct { |
| 243 | #ifdef __BIG_ENDIAN |
| 244 | u32 high, low; |
| 245 | #else |
| 246 | u32 low, high; |
| 247 | #endif |
| 248 | } l; |
| 249 | } u, rl, rh; |
| 250 | |
| 251 | u.ll = a; |
| 252 | rl.ll = mul_u32_u32(u.l.low, mul); |
| 253 | rh.ll = mul_u32_u32(u.l.high, mul) + rl.l.high; |
| 254 | |
| 255 | /* Bits 32-63 of the result will be in rh.l.low. */ |
| 256 | rl.l.high = do_div(rh.ll, divisor); |
| 257 | |
| 258 | /* Bits 0-31 of the result will be in rl.l.low. */ |
| 259 | do_div(rl.ll, divisor); |
| 260 | |
| 261 | rl.l.high = rh.l.low; |
| 262 | return rl.ll; |
| 263 | } |
| 264 | #endif /* mul_u64_u32_div */ |
| 265 | |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 266 | u64 mul_u64_u64_div_u64(u64 a, u64 mul, u64 div); |
| 267 | |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 268 | #define DIV64_U64_ROUND_UP(ll, d) \ |
| 269 | ({ u64 _tmp = (d); div64_u64((ll) + _tmp - 1, _tmp); }) |
| 270 | |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame] | 271 | /** |
| 272 | * DIV64_U64_ROUND_CLOSEST - unsigned 64bit divide with 64bit divisor rounded to nearest integer |
| 273 | * @dividend: unsigned 64bit dividend |
| 274 | * @divisor: unsigned 64bit divisor |
| 275 | * |
| 276 | * Divide unsigned 64bit dividend by unsigned 64bit divisor |
| 277 | * and round to closest integer. |
| 278 | * |
| 279 | * Return: dividend / divisor rounded to nearest integer |
| 280 | */ |
| 281 | #define DIV64_U64_ROUND_CLOSEST(dividend, divisor) \ |
| 282 | ({ u64 _tmp = (divisor); div64_u64((dividend) + _tmp / 2, _tmp); }) |
| 283 | |
Olivier Deprez | 157378f | 2022-04-04 15:47:50 +0200 | [diff] [blame] | 284 | /* |
| 285 | * DIV_S64_ROUND_CLOSEST - signed 64bit divide with 32bit divisor rounded to nearest integer |
| 286 | * @dividend: signed 64bit dividend |
| 287 | * @divisor: signed 32bit divisor |
| 288 | * |
| 289 | * Divide signed 64bit dividend by signed 32bit divisor |
| 290 | * and round to closest integer. |
| 291 | * |
| 292 | * Return: dividend / divisor rounded to nearest integer |
| 293 | */ |
| 294 | #define DIV_S64_ROUND_CLOSEST(dividend, divisor)( \ |
| 295 | { \ |
| 296 | s64 __x = (dividend); \ |
| 297 | s32 __d = (divisor); \ |
| 298 | ((__x > 0) == (__d > 0)) ? \ |
| 299 | div_s64((__x + (__d / 2)), __d) : \ |
| 300 | div_s64((__x - (__d / 2)), __d); \ |
| 301 | } \ |
| 302 | ) |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 303 | #endif /* _LINUX_MATH64_H */ |