Fred Reitberger | 2a099f1 | 2022-09-01 14:58:44 -0400 | [diff] [blame] | 1 | /* SPDX-License-Identifier: BSD-2-Clause */ |
| 2 | |
| 3 | /* |
| 4 | * xxHash - Extremely Fast Hash algorithm |
| 5 | * Copyright (C) 2012-2016, Yann Collet. |
| 6 | * |
| 7 | * You can contact the author at: |
| 8 | * - xxHash homepage: http://cyan4973.github.io/xxHash/ |
| 9 | * - xxHash source repository: https://github.com/Cyan4973/xxHash |
| 10 | */ |
| 11 | |
| 12 | #include <arch/byteorder.h> |
| 13 | #include <endian.h> |
| 14 | #include <string.h> |
| 15 | #include <xxhash.h> |
| 16 | |
| 17 | /*-************************************* |
| 18 | * Macros |
| 19 | **************************************/ |
| 20 | #define xxh_rotl32(x, r) ((x << r) | (x >> (32 - r))) |
| 21 | #define xxh_rotl64(x, r) ((x << r) | (x >> (64 - r))) |
| 22 | |
| 23 | /*-************************************* |
| 24 | * Constants |
| 25 | **************************************/ |
| 26 | static const uint32_t PRIME32_1 = 2654435761U; |
| 27 | static const uint32_t PRIME32_2 = 2246822519U; |
| 28 | static const uint32_t PRIME32_3 = 3266489917U; |
| 29 | static const uint32_t PRIME32_4 = 668265263U; |
| 30 | static const uint32_t PRIME32_5 = 374761393U; |
| 31 | |
| 32 | static const uint64_t PRIME64_1 = 11400714785074694791ULL; |
| 33 | static const uint64_t PRIME64_2 = 14029467366897019727ULL; |
| 34 | static const uint64_t PRIME64_3 = 1609587929392839161ULL; |
| 35 | static const uint64_t PRIME64_4 = 9650029242287828579ULL; |
| 36 | static const uint64_t PRIME64_5 = 2870177450012600261ULL; |
| 37 | |
| 38 | /*-************************** |
| 39 | * Utils |
| 40 | ***************************/ |
| 41 | void xxh32_copy_state(struct xxh32_state *dst, const struct xxh32_state *src) |
| 42 | { |
| 43 | memcpy(dst, src, sizeof(*dst)); |
| 44 | } |
| 45 | |
| 46 | void xxh64_copy_state(struct xxh64_state *dst, const struct xxh64_state *src) |
| 47 | { |
| 48 | memcpy(dst, src, sizeof(*dst)); |
| 49 | } |
| 50 | |
| 51 | static uint32_t xxh_get_unaligned_le32(const void *p) |
| 52 | { |
| 53 | const uint32_t *p32 = (const uint32_t *)p; |
| 54 | return le32toh(*p32); |
| 55 | } |
| 56 | |
| 57 | static uint64_t xxh_get_unaligned_le64(const void *p) |
| 58 | { |
| 59 | const uint64_t *p64 = (const uint64_t *)p; |
| 60 | return le64toh(*p64); |
| 61 | } |
| 62 | |
| 63 | /*-*************************** |
| 64 | * Simple Hash Functions |
| 65 | ****************************/ |
| 66 | static uint32_t xxh32_round(uint32_t seed, const uint32_t input) |
| 67 | { |
| 68 | seed += input * PRIME32_2; |
| 69 | seed = xxh_rotl32(seed, 13); |
| 70 | seed *= PRIME32_1; |
| 71 | return seed; |
| 72 | } |
| 73 | |
| 74 | uint32_t xxh32(const void *input, const size_t len, const uint32_t seed) |
| 75 | { |
| 76 | const uint8_t *p = (const uint8_t *)input; |
| 77 | const uint8_t *b_end = p + len; |
| 78 | uint32_t h32; |
| 79 | |
| 80 | if (len >= 16) { |
| 81 | const uint8_t *const limit = b_end - 16; |
| 82 | uint32_t v1 = seed + PRIME32_1 + PRIME32_2; |
| 83 | uint32_t v2 = seed + PRIME32_2; |
| 84 | uint32_t v3 = seed + 0; |
| 85 | uint32_t v4 = seed - PRIME32_1; |
| 86 | |
| 87 | do { |
| 88 | v1 = xxh32_round(v1, xxh_get_unaligned_le32(p)); |
| 89 | p += 4; |
| 90 | v2 = xxh32_round(v2, xxh_get_unaligned_le32(p)); |
| 91 | p += 4; |
| 92 | v3 = xxh32_round(v3, xxh_get_unaligned_le32(p)); |
| 93 | p += 4; |
| 94 | v4 = xxh32_round(v4, xxh_get_unaligned_le32(p)); |
| 95 | p += 4; |
| 96 | } while (p <= limit); |
| 97 | |
| 98 | h32 = xxh_rotl32(v1, 1) + xxh_rotl32(v2, 7) + |
| 99 | xxh_rotl32(v3, 12) + xxh_rotl32(v4, 18); |
| 100 | } else { |
| 101 | h32 = seed + PRIME32_5; |
| 102 | } |
| 103 | |
| 104 | h32 += (uint32_t)len; |
| 105 | |
| 106 | while (p + 4 <= b_end) { |
| 107 | h32 += xxh_get_unaligned_le32(p) * PRIME32_3; |
| 108 | h32 = xxh_rotl32(h32, 17) * PRIME32_4; |
| 109 | p += 4; |
| 110 | } |
| 111 | |
| 112 | while (p < b_end) { |
| 113 | h32 += (*p) * PRIME32_5; |
| 114 | h32 = xxh_rotl32(h32, 11) * PRIME32_1; |
| 115 | p++; |
| 116 | } |
| 117 | |
| 118 | h32 ^= h32 >> 15; |
| 119 | h32 *= PRIME32_2; |
| 120 | h32 ^= h32 >> 13; |
| 121 | h32 *= PRIME32_3; |
| 122 | h32 ^= h32 >> 16; |
| 123 | |
| 124 | return h32; |
| 125 | } |
| 126 | |
| 127 | static uint64_t xxh64_round(uint64_t acc, const uint64_t input) |
| 128 | { |
| 129 | acc += input * PRIME64_2; |
| 130 | acc = xxh_rotl64(acc, 31); |
| 131 | acc *= PRIME64_1; |
| 132 | return acc; |
| 133 | } |
| 134 | |
| 135 | static uint64_t xxh64_merge_round(uint64_t acc, uint64_t val) |
| 136 | { |
| 137 | val = xxh64_round(0, val); |
| 138 | acc ^= val; |
| 139 | acc = acc * PRIME64_1 + PRIME64_4; |
| 140 | return acc; |
| 141 | } |
| 142 | |
| 143 | uint64_t xxh64(const void *input, const size_t len, const uint64_t seed) |
| 144 | { |
| 145 | const uint8_t *p = (const uint8_t *)input; |
| 146 | const uint8_t *const b_end = p + len; |
| 147 | uint64_t h64; |
| 148 | |
| 149 | if (len >= 32) { |
| 150 | const uint8_t *const limit = b_end - 32; |
| 151 | uint64_t v1 = seed + PRIME64_1 + PRIME64_2; |
| 152 | uint64_t v2 = seed + PRIME64_2; |
| 153 | uint64_t v3 = seed + 0; |
| 154 | uint64_t v4 = seed - PRIME64_1; |
| 155 | |
| 156 | do { |
| 157 | v1 = xxh64_round(v1, xxh_get_unaligned_le64(p)); |
| 158 | p += 8; |
| 159 | v2 = xxh64_round(v2, xxh_get_unaligned_le64(p)); |
| 160 | p += 8; |
| 161 | v3 = xxh64_round(v3, xxh_get_unaligned_le64(p)); |
| 162 | p += 8; |
| 163 | v4 = xxh64_round(v4, xxh_get_unaligned_le64(p)); |
| 164 | p += 8; |
| 165 | } while (p <= limit); |
| 166 | |
| 167 | h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) + |
| 168 | xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18); |
| 169 | h64 = xxh64_merge_round(h64, v1); |
| 170 | h64 = xxh64_merge_round(h64, v2); |
| 171 | h64 = xxh64_merge_round(h64, v3); |
| 172 | h64 = xxh64_merge_round(h64, v4); |
| 173 | |
| 174 | } else { |
| 175 | h64 = seed + PRIME64_5; |
| 176 | } |
| 177 | |
| 178 | h64 += (uint64_t)len; |
| 179 | |
| 180 | while (p + 8 <= b_end) { |
| 181 | const uint64_t k1 = xxh64_round(0, xxh_get_unaligned_le64(p)); |
| 182 | |
| 183 | h64 ^= k1; |
| 184 | h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4; |
| 185 | p += 8; |
| 186 | } |
| 187 | |
| 188 | if (p + 4 <= b_end) { |
| 189 | h64 ^= (uint64_t)(xxh_get_unaligned_le32(p)) * PRIME64_1; |
| 190 | h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; |
| 191 | p += 4; |
| 192 | } |
| 193 | |
| 194 | while (p < b_end) { |
| 195 | h64 ^= (*p) * PRIME64_5; |
| 196 | h64 = xxh_rotl64(h64, 11) * PRIME64_1; |
| 197 | p++; |
| 198 | } |
| 199 | |
| 200 | h64 ^= h64 >> 33; |
| 201 | h64 *= PRIME64_2; |
| 202 | h64 ^= h64 >> 29; |
| 203 | h64 *= PRIME64_3; |
| 204 | h64 ^= h64 >> 32; |
| 205 | |
| 206 | return h64; |
| 207 | } |
| 208 | |
| 209 | /*-************************************************** |
| 210 | * Advanced Hash Functions |
| 211 | ***************************************************/ |
| 212 | void xxh32_reset(struct xxh32_state *statePtr, const uint32_t seed) |
| 213 | { |
| 214 | /* use a local state for memcpy() to avoid strict-aliasing warnings */ |
| 215 | struct xxh32_state state; |
| 216 | |
| 217 | memset(&state, 0, sizeof(state)); |
| 218 | state.v1 = seed + PRIME32_1 + PRIME32_2; |
| 219 | state.v2 = seed + PRIME32_2; |
| 220 | state.v3 = seed + 0; |
| 221 | state.v4 = seed - PRIME32_1; |
| 222 | memcpy(statePtr, &state, sizeof(state)); |
| 223 | } |
| 224 | |
| 225 | void xxh64_reset(struct xxh64_state *statePtr, const uint64_t seed) |
| 226 | { |
| 227 | /* use a local state for memcpy() to avoid strict-aliasing warnings */ |
| 228 | struct xxh64_state state; |
| 229 | |
| 230 | memset(&state, 0, sizeof(state)); |
| 231 | state.v1 = seed + PRIME64_1 + PRIME64_2; |
| 232 | state.v2 = seed + PRIME64_2; |
| 233 | state.v3 = seed + 0; |
| 234 | state.v4 = seed - PRIME64_1; |
| 235 | memcpy(statePtr, &state, sizeof(state)); |
| 236 | } |
| 237 | |
| 238 | int xxh32_update(struct xxh32_state *state, const void *input, const size_t len) |
| 239 | { |
| 240 | const uint8_t *p = (const uint8_t *)input; |
| 241 | const uint8_t *const b_end = p + len; |
| 242 | |
| 243 | if (input == NULL) |
| 244 | return -1; |
| 245 | |
| 246 | state->total_len_32 += (uint32_t)len; |
| 247 | state->large_len |= (len >= 16) | (state->total_len_32 >= 16); |
| 248 | |
| 249 | if (state->memsize + len < 16) { /* fill in tmp buffer */ |
| 250 | memcpy((uint8_t *)(state->mem32) + state->memsize, input, len); |
| 251 | state->memsize += (uint32_t)len; |
| 252 | return 0; |
| 253 | } |
| 254 | |
| 255 | if (state->memsize) { /* some data left from previous update */ |
| 256 | const uint32_t *p32 = state->mem32; |
| 257 | |
| 258 | memcpy((uint8_t *)(state->mem32) + state->memsize, input, |
| 259 | 16 - state->memsize); |
| 260 | |
| 261 | state->v1 = xxh32_round(state->v1, xxh_get_unaligned_le32(p32)); |
| 262 | p32++; |
| 263 | state->v2 = xxh32_round(state->v2, xxh_get_unaligned_le32(p32)); |
| 264 | p32++; |
| 265 | state->v3 = xxh32_round(state->v3, xxh_get_unaligned_le32(p32)); |
| 266 | p32++; |
| 267 | state->v4 = xxh32_round(state->v4, xxh_get_unaligned_le32(p32)); |
| 268 | p32++; |
| 269 | |
| 270 | p += 16-state->memsize; |
| 271 | state->memsize = 0; |
| 272 | } |
| 273 | |
| 274 | if (p <= b_end - 16) { |
| 275 | const uint8_t *const limit = b_end - 16; |
| 276 | uint32_t v1 = state->v1; |
| 277 | uint32_t v2 = state->v2; |
| 278 | uint32_t v3 = state->v3; |
| 279 | uint32_t v4 = state->v4; |
| 280 | |
| 281 | do { |
| 282 | v1 = xxh32_round(v1, xxh_get_unaligned_le32(p)); |
| 283 | p += 4; |
| 284 | v2 = xxh32_round(v2, xxh_get_unaligned_le32(p)); |
| 285 | p += 4; |
| 286 | v3 = xxh32_round(v3, xxh_get_unaligned_le32(p)); |
| 287 | p += 4; |
| 288 | v4 = xxh32_round(v4, xxh_get_unaligned_le32(p)); |
| 289 | p += 4; |
| 290 | } while (p <= limit); |
| 291 | |
| 292 | state->v1 = v1; |
| 293 | state->v2 = v2; |
| 294 | state->v3 = v3; |
| 295 | state->v4 = v4; |
| 296 | } |
| 297 | |
| 298 | if (p < b_end) { |
| 299 | memcpy(state->mem32, p, (size_t)(b_end-p)); |
| 300 | state->memsize = (uint32_t)(b_end-p); |
| 301 | } |
| 302 | |
| 303 | return 0; |
| 304 | } |
| 305 | |
| 306 | uint32_t xxh32_digest(const struct xxh32_state *state) |
| 307 | { |
| 308 | const uint8_t *p = (const uint8_t *)state->mem32; |
| 309 | const uint8_t *const b_end = (const uint8_t *)(state->mem32) + |
| 310 | state->memsize; |
| 311 | uint32_t h32; |
| 312 | |
| 313 | if (state->large_len) { |
| 314 | h32 = xxh_rotl32(state->v1, 1) + xxh_rotl32(state->v2, 7) + |
| 315 | xxh_rotl32(state->v3, 12) + xxh_rotl32(state->v4, 18); |
| 316 | } else { |
| 317 | h32 = state->v3 /* == seed */ + PRIME32_5; |
| 318 | } |
| 319 | |
| 320 | h32 += state->total_len_32; |
| 321 | |
| 322 | while (p + 4 <= b_end) { |
| 323 | h32 += xxh_get_unaligned_le32(p) * PRIME32_3; |
| 324 | h32 = xxh_rotl32(h32, 17) * PRIME32_4; |
| 325 | p += 4; |
| 326 | } |
| 327 | |
| 328 | while (p < b_end) { |
| 329 | h32 += (*p) * PRIME32_5; |
| 330 | h32 = xxh_rotl32(h32, 11) * PRIME32_1; |
| 331 | p++; |
| 332 | } |
| 333 | |
| 334 | h32 ^= h32 >> 15; |
| 335 | h32 *= PRIME32_2; |
| 336 | h32 ^= h32 >> 13; |
| 337 | h32 *= PRIME32_3; |
| 338 | h32 ^= h32 >> 16; |
| 339 | |
| 340 | return h32; |
| 341 | } |
| 342 | |
| 343 | int xxh64_update(struct xxh64_state *state, const void *input, const size_t len) |
| 344 | { |
| 345 | const uint8_t *p = (const uint8_t *)input; |
| 346 | const uint8_t *const b_end = p + len; |
| 347 | |
| 348 | if (input == NULL) |
| 349 | return -1; |
| 350 | |
| 351 | state->total_len += len; |
| 352 | |
| 353 | if (state->memsize + len < 32) { /* fill in tmp buffer */ |
| 354 | memcpy(((uint8_t *)state->mem64) + state->memsize, input, len); |
| 355 | state->memsize += (uint32_t)len; |
| 356 | return 0; |
| 357 | } |
| 358 | |
| 359 | if (state->memsize) { /* tmp buffer is full */ |
| 360 | uint64_t *p64 = state->mem64; |
| 361 | |
| 362 | memcpy(((uint8_t *)p64) + state->memsize, input, |
| 363 | 32 - state->memsize); |
| 364 | |
| 365 | state->v1 = xxh64_round(state->v1, xxh_get_unaligned_le64(p64)); |
| 366 | p64++; |
| 367 | state->v2 = xxh64_round(state->v2, xxh_get_unaligned_le64(p64)); |
| 368 | p64++; |
| 369 | state->v3 = xxh64_round(state->v3, xxh_get_unaligned_le64(p64)); |
| 370 | p64++; |
| 371 | state->v4 = xxh64_round(state->v4, xxh_get_unaligned_le64(p64)); |
| 372 | |
| 373 | p += 32 - state->memsize; |
| 374 | state->memsize = 0; |
| 375 | } |
| 376 | |
| 377 | if (p + 32 <= b_end) { |
| 378 | const uint8_t *const limit = b_end - 32; |
| 379 | uint64_t v1 = state->v1; |
| 380 | uint64_t v2 = state->v2; |
| 381 | uint64_t v3 = state->v3; |
| 382 | uint64_t v4 = state->v4; |
| 383 | |
| 384 | do { |
| 385 | v1 = xxh64_round(v1, xxh_get_unaligned_le64(p)); |
| 386 | p += 8; |
| 387 | v2 = xxh64_round(v2, xxh_get_unaligned_le64(p)); |
| 388 | p += 8; |
| 389 | v3 = xxh64_round(v3, xxh_get_unaligned_le64(p)); |
| 390 | p += 8; |
| 391 | v4 = xxh64_round(v4, xxh_get_unaligned_le64(p)); |
| 392 | p += 8; |
| 393 | } while (p <= limit); |
| 394 | |
| 395 | state->v1 = v1; |
| 396 | state->v2 = v2; |
| 397 | state->v3 = v3; |
| 398 | state->v4 = v4; |
| 399 | } |
| 400 | |
| 401 | if (p < b_end) { |
| 402 | memcpy(state->mem64, p, (size_t)(b_end-p)); |
| 403 | state->memsize = (uint32_t)(b_end - p); |
| 404 | } |
| 405 | |
| 406 | return 0; |
| 407 | } |
| 408 | |
| 409 | uint64_t xxh64_digest(const struct xxh64_state *state) |
| 410 | { |
| 411 | const uint8_t *p = (const uint8_t *)state->mem64; |
| 412 | const uint8_t *const b_end = (const uint8_t *)state->mem64 + |
| 413 | state->memsize; |
| 414 | uint64_t h64; |
| 415 | |
| 416 | if (state->total_len >= 32) { |
| 417 | const uint64_t v1 = state->v1; |
| 418 | const uint64_t v2 = state->v2; |
| 419 | const uint64_t v3 = state->v3; |
| 420 | const uint64_t v4 = state->v4; |
| 421 | |
| 422 | h64 = xxh_rotl64(v1, 1) + xxh_rotl64(v2, 7) + |
| 423 | xxh_rotl64(v3, 12) + xxh_rotl64(v4, 18); |
| 424 | h64 = xxh64_merge_round(h64, v1); |
| 425 | h64 = xxh64_merge_round(h64, v2); |
| 426 | h64 = xxh64_merge_round(h64, v3); |
| 427 | h64 = xxh64_merge_round(h64, v4); |
| 428 | } else { |
| 429 | h64 = state->v3 + PRIME64_5; |
| 430 | } |
| 431 | |
| 432 | h64 += (uint64_t)state->total_len; |
| 433 | |
| 434 | while (p + 8 <= b_end) { |
| 435 | const uint64_t k1 = xxh64_round(0, xxh_get_unaligned_le64(p)); |
| 436 | |
| 437 | h64 ^= k1; |
| 438 | h64 = xxh_rotl64(h64, 27) * PRIME64_1 + PRIME64_4; |
| 439 | p += 8; |
| 440 | } |
| 441 | |
| 442 | if (p + 4 <= b_end) { |
| 443 | h64 ^= (uint64_t)(xxh_get_unaligned_le32(p)) * PRIME64_1; |
| 444 | h64 = xxh_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; |
| 445 | p += 4; |
| 446 | } |
| 447 | |
| 448 | while (p < b_end) { |
| 449 | h64 ^= (*p) * PRIME64_5; |
| 450 | h64 = xxh_rotl64(h64, 11) * PRIME64_1; |
| 451 | p++; |
| 452 | } |
| 453 | |
| 454 | h64 ^= h64 >> 33; |
| 455 | h64 *= PRIME64_2; |
| 456 | h64 ^= h64 >> 29; |
| 457 | h64 *= PRIME64_3; |
| 458 | h64 ^= h64 >> 32; |
| 459 | |
| 460 | return h64; |
| 461 | } |