Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 1 | /* |
| 2 | xxHash - Fast Hash algorithm |
| 3 | Copyright (C) 2012-2014, Yann Collet. |
| 4 | BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php) |
| 5 | |
| 6 | Redistribution and use in source and binary forms, with or without |
| 7 | modification, are permitted provided that the following conditions are |
| 8 | met: |
| 9 | |
| 10 | * Redistributions of source code must retain the above copyright |
| 11 | notice, this list of conditions and the following disclaimer. |
| 12 | * Redistributions in binary form must reproduce the above |
| 13 | copyright notice, this list of conditions and the following disclaimer |
| 14 | in the documentation and/or other materials provided with the |
| 15 | distribution. |
| 16 | |
| 17 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 18 | "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 19 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 20 | A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 21 | OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 22 | SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 23 | LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 24 | DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 25 | THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 26 | (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 27 | OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 28 | |
| 29 | You can contact the author at : |
| 30 | - xxHash source repository : http://code.google.com/p/xxhash/ |
| 31 | - public discussion board : https://groups.google.com/forum/#!forum/lz4c |
| 32 | */ |
| 33 | |
| 34 | |
| 35 | //************************************** |
| 36 | // Tuning parameters |
| 37 | //************************************** |
| 38 | // Unaligned memory access is automatically enabled for "common" CPU, such as x86. |
| 39 | // For others CPU, the compiler will be more cautious, and insert extra code to ensure aligned access is respected. |
| 40 | // If you know your target CPU supports unaligned memory access, you want to force this option manually to improve performance. |
| 41 | // You can also enable this parameter if you know your input data will always be aligned (boundaries of 4, for U32). |
Willy Tarreau | 59ad20e | 2021-02-04 17:02:39 +0100 | [diff] [blame] | 42 | // 32-bit ARM is more annoying, modern cores do support unaligned accesses, but |
| 43 | // not on 64-bit data (the ldrd instructions causes an alignment exception). |
| 44 | // Because of this we need to split the condition for 32 and 64 bit. |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 45 | #if defined(__ARM_FEATURE_UNALIGNED) || defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64) |
| 46 | # define XXH_USE_UNALIGNED_ACCESS 1 |
Willy Tarreau | 59ad20e | 2021-02-04 17:02:39 +0100 | [diff] [blame] | 47 | # if !defined(__arm__) |
| 48 | # define XXH_USE_UNALIGNED_ACCESS64 1 |
| 49 | # endif |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 50 | #endif |
| 51 | |
| 52 | // XXH_ACCEPT_NULL_INPUT_POINTER : |
| 53 | // If the input pointer is a null pointer, xxHash default behavior is to trigger a memory access error, since it is a bad pointer. |
| 54 | // When this option is enabled, xxHash output for null input pointers will be the same as a null-length input. |
| 55 | // This option has a very small performance cost (only measurable on small inputs). |
| 56 | // By default, this option is disabled. To enable it, uncomment below define : |
| 57 | // #define XXH_ACCEPT_NULL_INPUT_POINTER 1 |
| 58 | |
| 59 | // XXH_FORCE_NATIVE_FORMAT : |
Joseph Herlant | b2db6a0 | 2018-11-15 09:30:49 -0800 | [diff] [blame] | 60 | // By default, xxHash library provides endian-independent Hash values, based on little-endian convention. |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 61 | // Results are therefore identical for little-endian and big-endian CPU. |
| 62 | // This comes at a performance cost for big-endian CPU, since some swapping is required to emulate little-endian format. |
Joseph Herlant | b2db6a0 | 2018-11-15 09:30:49 -0800 | [diff] [blame] | 63 | // Should endian-independence be of no importance for your application, you may set the #define below to 1. |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 64 | // It will improve speed for Big-endian CPU. |
| 65 | // This option has no impact on Little_Endian CPU. |
| 66 | #define XXH_FORCE_NATIVE_FORMAT 0 |
| 67 | |
| 68 | //************************************** |
| 69 | // Compiler Specific Options |
| 70 | //************************************** |
| 71 | // Disable some Visual warning messages |
| 72 | #ifdef _MSC_VER // Visual Studio |
| 73 | # pragma warning(disable : 4127) // disable: C4127: conditional expression is constant |
| 74 | #endif |
| 75 | |
| 76 | #ifdef _MSC_VER // Visual Studio |
| 77 | # define FORCE_INLINE static __forceinline |
| 78 | #else |
| 79 | # ifdef __GNUC__ |
| 80 | # define FORCE_INLINE static inline __attribute__((always_inline)) |
| 81 | # else |
| 82 | # define FORCE_INLINE static inline |
| 83 | # endif |
| 84 | #endif |
| 85 | |
| 86 | //************************************** |
| 87 | // Includes & Memory related functions |
| 88 | //************************************** |
| 89 | #include <import/xxhash.h> |
| 90 | // Modify the local functions below should you wish to use some other memory routines |
| 91 | // for malloc(), free() |
| 92 | #include <stdlib.h> |
| 93 | static void* XXH_malloc(size_t s) { return malloc(s); } |
| 94 | static void XXH_free (void* p) { free(p); } |
| 95 | // for memcpy() |
| 96 | #include <string.h> |
| 97 | static void* XXH_memcpy(void* dest, const void* src, size_t size) |
| 98 | { |
| 99 | return memcpy(dest,src,size); |
| 100 | } |
| 101 | |
| 102 | |
| 103 | //************************************** |
| 104 | // Basic Types |
| 105 | //************************************** |
| 106 | #if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L // C99 |
Willy Tarreau | a1bd1fa | 2019-03-29 17:26:33 +0100 | [diff] [blame] | 107 | # include <inttypes.h> |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 108 | typedef uint8_t BYTE; |
| 109 | typedef uint16_t U16; |
| 110 | typedef uint32_t U32; |
| 111 | typedef int32_t S32; |
| 112 | typedef uint64_t U64; |
| 113 | #else |
| 114 | typedef unsigned char BYTE; |
| 115 | typedef unsigned short U16; |
| 116 | typedef unsigned int U32; |
| 117 | typedef signed int S32; |
| 118 | typedef unsigned long long U64; |
| 119 | #endif |
| 120 | |
| 121 | #if defined(__GNUC__) && !defined(XXH_USE_UNALIGNED_ACCESS) |
| 122 | # define _PACKED __attribute__ ((packed)) |
| 123 | #else |
| 124 | # define _PACKED |
| 125 | #endif |
| 126 | |
Willy Tarreau | 59ad20e | 2021-02-04 17:02:39 +0100 | [diff] [blame] | 127 | #if defined(__GNUC__) && !defined(XXH_USE_UNALIGNED_ACCESS64) |
| 128 | # define _PACKED64 __attribute__ ((packed)) |
| 129 | #else |
| 130 | # define _PACKED64 |
| 131 | #endif |
| 132 | |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 133 | #if !defined(XXH_USE_UNALIGNED_ACCESS) && !defined(__GNUC__) |
| 134 | # ifdef __IBMC__ |
| 135 | # pragma pack(1) |
| 136 | # else |
| 137 | # pragma pack(push, 1) |
| 138 | # endif |
| 139 | #endif |
| 140 | |
| 141 | typedef struct _U32_S |
| 142 | { |
| 143 | U32 v; |
| 144 | } _PACKED U32_S; |
| 145 | typedef struct _U64_S |
| 146 | { |
| 147 | U64 v; |
Willy Tarreau | 59ad20e | 2021-02-04 17:02:39 +0100 | [diff] [blame] | 148 | } _PACKED64 U64_S; |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 149 | |
| 150 | #if !defined(XXH_USE_UNALIGNED_ACCESS) && !defined(__GNUC__) |
| 151 | # pragma pack(pop) |
| 152 | #endif |
| 153 | |
| 154 | #define A32(x) (((U32_S *)(x))->v) |
| 155 | #define A64(x) (((U64_S *)(x))->v) |
| 156 | |
| 157 | |
| 158 | //*************************************** |
| 159 | // Compiler-specific Functions and Macros |
| 160 | //*************************************** |
| 161 | #define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__) |
| 162 | |
| 163 | // Note : although _rotl exists for minGW (GCC under windows), performance seems poor |
| 164 | #if defined(_MSC_VER) |
| 165 | # define XXH_rotl32(x,r) _rotl(x,r) |
| 166 | # define XXH_rotl64(x,r) _rotl64(x,r) |
| 167 | #else |
| 168 | # define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r))) |
| 169 | # define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r))) |
| 170 | #endif |
| 171 | |
| 172 | #if defined(_MSC_VER) // Visual Studio |
| 173 | # define XXH_swap32 _byteswap_ulong |
| 174 | # define XXH_swap64 _byteswap_uint64 |
| 175 | #elif GCC_VERSION >= 403 |
| 176 | # define XXH_swap32 __builtin_bswap32 |
| 177 | # define XXH_swap64 __builtin_bswap64 |
| 178 | #else |
| 179 | static inline U32 XXH_swap32 (U32 x) |
| 180 | { |
| 181 | return ((x << 24) & 0xff000000 ) | |
| 182 | ((x << 8) & 0x00ff0000 ) | |
| 183 | ((x >> 8) & 0x0000ff00 ) | |
| 184 | ((x >> 24) & 0x000000ff ); |
| 185 | } |
| 186 | static inline U64 XXH_swap64 (U64 x) |
| 187 | { |
| 188 | return ((x << 56) & 0xff00000000000000ULL) | |
| 189 | ((x << 40) & 0x00ff000000000000ULL) | |
| 190 | ((x << 24) & 0x0000ff0000000000ULL) | |
| 191 | ((x << 8) & 0x000000ff00000000ULL) | |
| 192 | ((x >> 8) & 0x00000000ff000000ULL) | |
| 193 | ((x >> 24) & 0x0000000000ff0000ULL) | |
| 194 | ((x >> 40) & 0x000000000000ff00ULL) | |
| 195 | ((x >> 56) & 0x00000000000000ffULL); |
| 196 | } |
| 197 | #endif |
| 198 | |
| 199 | |
| 200 | //************************************** |
| 201 | // Constants |
| 202 | //************************************** |
| 203 | #define PRIME32_1 2654435761U |
| 204 | #define PRIME32_2 2246822519U |
| 205 | #define PRIME32_3 3266489917U |
| 206 | #define PRIME32_4 668265263U |
| 207 | #define PRIME32_5 374761393U |
| 208 | |
| 209 | #define PRIME64_1 11400714785074694791ULL |
| 210 | #define PRIME64_2 14029467366897019727ULL |
| 211 | #define PRIME64_3 1609587929392839161ULL |
| 212 | #define PRIME64_4 9650029242287828579ULL |
| 213 | #define PRIME64_5 2870177450012600261ULL |
| 214 | |
| 215 | //************************************** |
| 216 | // Architecture Macros |
| 217 | //************************************** |
| 218 | typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess; |
| 219 | #ifndef XXH_CPU_LITTLE_ENDIAN // It is possible to define XXH_CPU_LITTLE_ENDIAN externally, for example using a compiler switch |
| 220 | static const int one = 1; |
| 221 | # define XXH_CPU_LITTLE_ENDIAN (*(char*)(&one)) |
| 222 | #endif |
| 223 | |
| 224 | |
| 225 | //************************************** |
| 226 | // Macros |
| 227 | //************************************** |
| 228 | #define XXH_STATIC_ASSERT(c) { enum { XXH_static_assert = 1/(!!(c)) }; } // use only *after* variable declarations |
| 229 | |
| 230 | |
| 231 | //**************************** |
| 232 | // Memory reads |
| 233 | //**************************** |
| 234 | typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment; |
| 235 | |
| 236 | FORCE_INLINE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align) |
| 237 | { |
| 238 | if (align==XXH_unaligned) |
| 239 | return endian==XXH_littleEndian ? A32(ptr) : XXH_swap32(A32(ptr)); |
| 240 | else |
| 241 | return endian==XXH_littleEndian ? *(U32*)ptr : XXH_swap32(*(U32*)ptr); |
| 242 | } |
| 243 | |
| 244 | FORCE_INLINE U32 XXH_readLE32(const void* ptr, XXH_endianess endian) |
| 245 | { |
| 246 | return XXH_readLE32_align(ptr, endian, XXH_unaligned); |
| 247 | } |
| 248 | |
| 249 | FORCE_INLINE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align) |
| 250 | { |
| 251 | if (align==XXH_unaligned) |
| 252 | return endian==XXH_littleEndian ? A64(ptr) : XXH_swap64(A64(ptr)); |
| 253 | else |
| 254 | return endian==XXH_littleEndian ? *(U64*)ptr : XXH_swap64(*(U64*)ptr); |
| 255 | } |
| 256 | |
| 257 | FORCE_INLINE U64 XXH_readLE64(const void* ptr, XXH_endianess endian) |
| 258 | { |
| 259 | return XXH_readLE64_align(ptr, endian, XXH_unaligned); |
| 260 | } |
| 261 | |
| 262 | |
| 263 | //**************************** |
| 264 | // Simple Hash Functions |
| 265 | //**************************** |
| 266 | FORCE_INLINE U32 XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_endianess endian, XXH_alignment align) |
| 267 | { |
Vincent Bernat | 3c2f2f2 | 2016-04-03 13:48:42 +0200 | [diff] [blame] | 268 | const BYTE* p = input; |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 269 | const BYTE* bEnd = p + len; |
| 270 | U32 h32; |
| 271 | #define XXH_get32bits(p) XXH_readLE32_align(p, endian, align) |
| 272 | |
| 273 | #ifdef XXH_ACCEPT_NULL_INPUT_POINTER |
| 274 | if (p==NULL) |
| 275 | { |
| 276 | len=0; |
| 277 | bEnd=p=(const BYTE*)(size_t)16; |
| 278 | } |
| 279 | #endif |
| 280 | |
| 281 | if (len>=16) |
| 282 | { |
| 283 | const BYTE* const limit = bEnd - 16; |
| 284 | U32 v1 = seed + PRIME32_1 + PRIME32_2; |
| 285 | U32 v2 = seed + PRIME32_2; |
| 286 | U32 v3 = seed + 0; |
| 287 | U32 v4 = seed - PRIME32_1; |
| 288 | |
| 289 | do |
| 290 | { |
| 291 | v1 += XXH_get32bits(p) * PRIME32_2; |
| 292 | v1 = XXH_rotl32(v1, 13); |
| 293 | v1 *= PRIME32_1; |
| 294 | p+=4; |
| 295 | v2 += XXH_get32bits(p) * PRIME32_2; |
| 296 | v2 = XXH_rotl32(v2, 13); |
| 297 | v2 *= PRIME32_1; |
| 298 | p+=4; |
| 299 | v3 += XXH_get32bits(p) * PRIME32_2; |
| 300 | v3 = XXH_rotl32(v3, 13); |
| 301 | v3 *= PRIME32_1; |
| 302 | p+=4; |
| 303 | v4 += XXH_get32bits(p) * PRIME32_2; |
| 304 | v4 = XXH_rotl32(v4, 13); |
| 305 | v4 *= PRIME32_1; |
| 306 | p+=4; |
| 307 | } |
| 308 | while (p<=limit); |
| 309 | |
| 310 | h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18); |
| 311 | } |
| 312 | else |
| 313 | { |
| 314 | h32 = seed + PRIME32_5; |
| 315 | } |
| 316 | |
| 317 | h32 += (U32) len; |
| 318 | |
| 319 | while (p+4<=bEnd) |
| 320 | { |
| 321 | h32 += XXH_get32bits(p) * PRIME32_3; |
| 322 | h32 = XXH_rotl32(h32, 17) * PRIME32_4 ; |
| 323 | p+=4; |
| 324 | } |
| 325 | |
| 326 | while (p<bEnd) |
| 327 | { |
| 328 | h32 += (*p) * PRIME32_5; |
| 329 | h32 = XXH_rotl32(h32, 11) * PRIME32_1 ; |
| 330 | p++; |
| 331 | } |
| 332 | |
| 333 | h32 ^= h32 >> 15; |
| 334 | h32 *= PRIME32_2; |
| 335 | h32 ^= h32 >> 13; |
| 336 | h32 *= PRIME32_3; |
| 337 | h32 ^= h32 >> 16; |
| 338 | |
| 339 | return h32; |
| 340 | } |
| 341 | |
| 342 | |
| 343 | unsigned int XXH32 (const void* input, size_t len, unsigned seed) |
| 344 | { |
| 345 | #if 0 |
| 346 | // Simple version, good for code maintenance, but unfortunately slow for small inputs |
| 347 | XXH32_state_t state; |
| 348 | XXH32_reset(&state, seed); |
| 349 | XXH32_update(&state, input, len); |
| 350 | return XXH32_digest(&state); |
| 351 | #else |
| 352 | XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; |
| 353 | |
| 354 | # if !defined(XXH_USE_UNALIGNED_ACCESS) |
| 355 | if ((((size_t)input) & 3) == 0) // Input is aligned, let's leverage the speed advantage |
| 356 | { |
| 357 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 358 | return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned); |
| 359 | else |
| 360 | return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned); |
| 361 | } |
| 362 | # endif |
| 363 | |
| 364 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 365 | return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned); |
| 366 | else |
| 367 | return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned); |
| 368 | #endif |
| 369 | } |
| 370 | |
| 371 | FORCE_INLINE U64 XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_endianess endian, XXH_alignment align) |
| 372 | { |
Vincent Bernat | 3c2f2f2 | 2016-04-03 13:48:42 +0200 | [diff] [blame] | 373 | const BYTE* p = input; |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 374 | const BYTE* bEnd = p + len; |
| 375 | U64 h64; |
| 376 | #define XXH_get64bits(p) XXH_readLE64_align(p, endian, align) |
| 377 | |
| 378 | #ifdef XXH_ACCEPT_NULL_INPUT_POINTER |
| 379 | if (p==NULL) |
| 380 | { |
| 381 | len=0; |
| 382 | bEnd=p=(const BYTE*)(size_t)32; |
| 383 | } |
| 384 | #endif |
| 385 | |
| 386 | if (len>=32) |
| 387 | { |
| 388 | const BYTE* const limit = bEnd - 32; |
| 389 | U64 v1 = seed + PRIME64_1 + PRIME64_2; |
| 390 | U64 v2 = seed + PRIME64_2; |
| 391 | U64 v3 = seed + 0; |
| 392 | U64 v4 = seed - PRIME64_1; |
| 393 | |
| 394 | do |
| 395 | { |
| 396 | v1 += XXH_get64bits(p) * PRIME64_2; |
| 397 | p+=8; |
| 398 | v1 = XXH_rotl64(v1, 31); |
| 399 | v1 *= PRIME64_1; |
| 400 | v2 += XXH_get64bits(p) * PRIME64_2; |
| 401 | p+=8; |
| 402 | v2 = XXH_rotl64(v2, 31); |
| 403 | v2 *= PRIME64_1; |
| 404 | v3 += XXH_get64bits(p) * PRIME64_2; |
| 405 | p+=8; |
| 406 | v3 = XXH_rotl64(v3, 31); |
| 407 | v3 *= PRIME64_1; |
| 408 | v4 += XXH_get64bits(p) * PRIME64_2; |
| 409 | p+=8; |
| 410 | v4 = XXH_rotl64(v4, 31); |
| 411 | v4 *= PRIME64_1; |
| 412 | } |
| 413 | while (p<=limit); |
| 414 | |
| 415 | h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18); |
| 416 | |
| 417 | v1 *= PRIME64_2; |
| 418 | v1 = XXH_rotl64(v1, 31); |
| 419 | v1 *= PRIME64_1; |
| 420 | h64 ^= v1; |
| 421 | h64 = h64 * PRIME64_1 + PRIME64_4; |
| 422 | |
| 423 | v2 *= PRIME64_2; |
| 424 | v2 = XXH_rotl64(v2, 31); |
| 425 | v2 *= PRIME64_1; |
| 426 | h64 ^= v2; |
| 427 | h64 = h64 * PRIME64_1 + PRIME64_4; |
| 428 | |
| 429 | v3 *= PRIME64_2; |
| 430 | v3 = XXH_rotl64(v3, 31); |
| 431 | v3 *= PRIME64_1; |
| 432 | h64 ^= v3; |
| 433 | h64 = h64 * PRIME64_1 + PRIME64_4; |
| 434 | |
| 435 | v4 *= PRIME64_2; |
| 436 | v4 = XXH_rotl64(v4, 31); |
| 437 | v4 *= PRIME64_1; |
| 438 | h64 ^= v4; |
| 439 | h64 = h64 * PRIME64_1 + PRIME64_4; |
| 440 | } |
| 441 | else |
| 442 | { |
| 443 | h64 = seed + PRIME64_5; |
| 444 | } |
| 445 | |
| 446 | h64 += (U64) len; |
| 447 | |
| 448 | while (p+8<=bEnd) |
| 449 | { |
| 450 | U64 k1 = XXH_get64bits(p); |
| 451 | k1 *= PRIME64_2; |
| 452 | k1 = XXH_rotl64(k1,31); |
| 453 | k1 *= PRIME64_1; |
| 454 | h64 ^= k1; |
| 455 | h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; |
| 456 | p+=8; |
| 457 | } |
| 458 | |
| 459 | if (p+4<=bEnd) |
| 460 | { |
| 461 | h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1; |
| 462 | h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; |
| 463 | p+=4; |
| 464 | } |
| 465 | |
| 466 | while (p<bEnd) |
| 467 | { |
| 468 | h64 ^= (*p) * PRIME64_5; |
| 469 | h64 = XXH_rotl64(h64, 11) * PRIME64_1; |
| 470 | p++; |
| 471 | } |
| 472 | |
| 473 | h64 ^= h64 >> 33; |
| 474 | h64 *= PRIME64_2; |
| 475 | h64 ^= h64 >> 29; |
| 476 | h64 *= PRIME64_3; |
| 477 | h64 ^= h64 >> 32; |
| 478 | |
| 479 | return h64; |
| 480 | } |
| 481 | |
| 482 | |
| 483 | unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed) |
| 484 | { |
| 485 | #if 0 |
| 486 | // Simple version, good for code maintenance, but unfortunately slow for small inputs |
| 487 | XXH64_state_t state; |
| 488 | XXH64_reset(&state, seed); |
| 489 | XXH64_update(&state, input, len); |
| 490 | return XXH64_digest(&state); |
| 491 | #else |
| 492 | XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; |
| 493 | |
Willy Tarreau | 59ad20e | 2021-02-04 17:02:39 +0100 | [diff] [blame] | 494 | # if !defined(XXH_USE_UNALIGNED_ACCESS64) |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 495 | if ((((size_t)input) & 7)==0) // Input is aligned, let's leverage the speed advantage |
| 496 | { |
| 497 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 498 | return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned); |
| 499 | else |
| 500 | return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned); |
| 501 | } |
| 502 | # endif |
| 503 | |
| 504 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 505 | return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned); |
| 506 | else |
| 507 | return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned); |
| 508 | #endif |
| 509 | } |
| 510 | |
| 511 | /**************************************************** |
| 512 | * Advanced Hash Functions |
| 513 | ****************************************************/ |
| 514 | |
| 515 | /*** Allocation ***/ |
| 516 | typedef struct |
| 517 | { |
| 518 | U64 total_len; |
| 519 | U32 seed; |
| 520 | U32 v1; |
| 521 | U32 v2; |
| 522 | U32 v3; |
| 523 | U32 v4; |
| 524 | U32 mem32[4]; /* defined as U32 for alignment */ |
| 525 | U32 memsize; |
| 526 | } XXH_istate32_t; |
| 527 | |
| 528 | typedef struct |
| 529 | { |
| 530 | U64 total_len; |
| 531 | U64 seed; |
| 532 | U64 v1; |
| 533 | U64 v2; |
| 534 | U64 v3; |
| 535 | U64 v4; |
| 536 | U64 mem64[4]; /* defined as U64 for alignment */ |
| 537 | U32 memsize; |
| 538 | } XXH_istate64_t; |
| 539 | |
| 540 | |
| 541 | XXH32_state_t* XXH32_createState(void) |
| 542 | { |
| 543 | XXH_STATIC_ASSERT(sizeof(XXH32_state_t) >= sizeof(XXH_istate32_t)); // A compilation error here means XXH32_state_t is not large enough |
| 544 | return (XXH32_state_t*)XXH_malloc(sizeof(XXH32_state_t)); |
| 545 | } |
| 546 | XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr) |
| 547 | { |
| 548 | XXH_free(statePtr); |
| 549 | return XXH_OK; |
| 550 | }; |
| 551 | |
| 552 | XXH64_state_t* XXH64_createState(void) |
| 553 | { |
| 554 | XXH_STATIC_ASSERT(sizeof(XXH64_state_t) >= sizeof(XXH_istate64_t)); // A compilation error here means XXH64_state_t is not large enough |
| 555 | return (XXH64_state_t*)XXH_malloc(sizeof(XXH64_state_t)); |
| 556 | } |
| 557 | XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr) |
| 558 | { |
| 559 | XXH_free(statePtr); |
| 560 | return XXH_OK; |
| 561 | }; |
| 562 | |
| 563 | |
| 564 | /*** Hash feed ***/ |
| 565 | |
| 566 | XXH_errorcode XXH32_reset(XXH32_state_t* state_in, U32 seed) |
| 567 | { |
| 568 | XXH_istate32_t* state = (XXH_istate32_t*) state_in; |
| 569 | state->seed = seed; |
| 570 | state->v1 = seed + PRIME32_1 + PRIME32_2; |
| 571 | state->v2 = seed + PRIME32_2; |
| 572 | state->v3 = seed + 0; |
| 573 | state->v4 = seed - PRIME32_1; |
| 574 | state->total_len = 0; |
| 575 | state->memsize = 0; |
| 576 | return XXH_OK; |
| 577 | } |
| 578 | |
| 579 | XXH_errorcode XXH64_reset(XXH64_state_t* state_in, unsigned long long seed) |
| 580 | { |
| 581 | XXH_istate64_t* state = (XXH_istate64_t*) state_in; |
| 582 | state->seed = seed; |
| 583 | state->v1 = seed + PRIME64_1 + PRIME64_2; |
| 584 | state->v2 = seed + PRIME64_2; |
| 585 | state->v3 = seed + 0; |
| 586 | state->v4 = seed - PRIME64_1; |
| 587 | state->total_len = 0; |
| 588 | state->memsize = 0; |
| 589 | return XXH_OK; |
| 590 | } |
| 591 | |
| 592 | |
| 593 | FORCE_INLINE XXH_errorcode XXH32_update_endian (XXH32_state_t* state_in, const void* input, size_t len, XXH_endianess endian) |
| 594 | { |
| 595 | XXH_istate32_t* state = (XXH_istate32_t *) state_in; |
Vincent Bernat | 3c2f2f2 | 2016-04-03 13:48:42 +0200 | [diff] [blame] | 596 | const BYTE* p = input; |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 597 | const BYTE* const bEnd = p + len; |
| 598 | |
| 599 | #ifdef XXH_ACCEPT_NULL_INPUT_POINTER |
| 600 | if (input==NULL) return XXH_ERROR; |
| 601 | #endif |
| 602 | |
| 603 | state->total_len += len; |
| 604 | |
| 605 | if (state->memsize + len < 16) // fill in tmp buffer |
| 606 | { |
| 607 | XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len); |
| 608 | state->memsize += (U32)len; |
| 609 | return XXH_OK; |
| 610 | } |
| 611 | |
| 612 | if (state->memsize) // some data left from previous update |
| 613 | { |
| 614 | XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize); |
| 615 | { |
| 616 | const U32* p32 = state->mem32; |
| 617 | state->v1 += XXH_readLE32(p32, endian) * PRIME32_2; |
| 618 | state->v1 = XXH_rotl32(state->v1, 13); |
| 619 | state->v1 *= PRIME32_1; |
| 620 | p32++; |
| 621 | state->v2 += XXH_readLE32(p32, endian) * PRIME32_2; |
| 622 | state->v2 = XXH_rotl32(state->v2, 13); |
| 623 | state->v2 *= PRIME32_1; |
| 624 | p32++; |
| 625 | state->v3 += XXH_readLE32(p32, endian) * PRIME32_2; |
| 626 | state->v3 = XXH_rotl32(state->v3, 13); |
| 627 | state->v3 *= PRIME32_1; |
| 628 | p32++; |
| 629 | state->v4 += XXH_readLE32(p32, endian) * PRIME32_2; |
| 630 | state->v4 = XXH_rotl32(state->v4, 13); |
| 631 | state->v4 *= PRIME32_1; |
| 632 | p32++; |
| 633 | } |
| 634 | p += 16-state->memsize; |
| 635 | state->memsize = 0; |
| 636 | } |
| 637 | |
| 638 | if (p <= bEnd-16) |
| 639 | { |
| 640 | const BYTE* const limit = bEnd - 16; |
| 641 | U32 v1 = state->v1; |
| 642 | U32 v2 = state->v2; |
| 643 | U32 v3 = state->v3; |
| 644 | U32 v4 = state->v4; |
| 645 | |
| 646 | do |
| 647 | { |
| 648 | v1 += XXH_readLE32(p, endian) * PRIME32_2; |
| 649 | v1 = XXH_rotl32(v1, 13); |
| 650 | v1 *= PRIME32_1; |
| 651 | p+=4; |
| 652 | v2 += XXH_readLE32(p, endian) * PRIME32_2; |
| 653 | v2 = XXH_rotl32(v2, 13); |
| 654 | v2 *= PRIME32_1; |
| 655 | p+=4; |
| 656 | v3 += XXH_readLE32(p, endian) * PRIME32_2; |
| 657 | v3 = XXH_rotl32(v3, 13); |
| 658 | v3 *= PRIME32_1; |
| 659 | p+=4; |
| 660 | v4 += XXH_readLE32(p, endian) * PRIME32_2; |
| 661 | v4 = XXH_rotl32(v4, 13); |
| 662 | v4 *= PRIME32_1; |
| 663 | p+=4; |
| 664 | } |
| 665 | while (p<=limit); |
| 666 | |
| 667 | state->v1 = v1; |
| 668 | state->v2 = v2; |
| 669 | state->v3 = v3; |
| 670 | state->v4 = v4; |
| 671 | } |
| 672 | |
| 673 | if (p < bEnd) |
| 674 | { |
| 675 | XXH_memcpy(state->mem32, p, bEnd-p); |
| 676 | state->memsize = (int)(bEnd-p); |
| 677 | } |
| 678 | |
| 679 | return XXH_OK; |
| 680 | } |
| 681 | |
| 682 | XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len) |
| 683 | { |
| 684 | XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; |
| 685 | |
| 686 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 687 | return XXH32_update_endian(state_in, input, len, XXH_littleEndian); |
| 688 | else |
| 689 | return XXH32_update_endian(state_in, input, len, XXH_bigEndian); |
| 690 | } |
| 691 | |
| 692 | |
| 693 | |
| 694 | FORCE_INLINE U32 XXH32_digest_endian (const XXH32_state_t* state_in, XXH_endianess endian) |
| 695 | { |
| 696 | XXH_istate32_t* state = (XXH_istate32_t*) state_in; |
| 697 | const BYTE * p = (const BYTE*)state->mem32; |
| 698 | BYTE* bEnd = (BYTE*)(state->mem32) + state->memsize; |
| 699 | U32 h32; |
| 700 | |
| 701 | if (state->total_len >= 16) |
| 702 | { |
| 703 | h32 = XXH_rotl32(state->v1, 1) + XXH_rotl32(state->v2, 7) + XXH_rotl32(state->v3, 12) + XXH_rotl32(state->v4, 18); |
| 704 | } |
| 705 | else |
| 706 | { |
| 707 | h32 = state->seed + PRIME32_5; |
| 708 | } |
| 709 | |
| 710 | h32 += (U32) state->total_len; |
| 711 | |
| 712 | while (p+4<=bEnd) |
| 713 | { |
| 714 | h32 += XXH_readLE32(p, endian) * PRIME32_3; |
| 715 | h32 = XXH_rotl32(h32, 17) * PRIME32_4; |
| 716 | p+=4; |
| 717 | } |
| 718 | |
| 719 | while (p<bEnd) |
| 720 | { |
| 721 | h32 += (*p) * PRIME32_5; |
| 722 | h32 = XXH_rotl32(h32, 11) * PRIME32_1; |
| 723 | p++; |
| 724 | } |
| 725 | |
| 726 | h32 ^= h32 >> 15; |
| 727 | h32 *= PRIME32_2; |
| 728 | h32 ^= h32 >> 13; |
| 729 | h32 *= PRIME32_3; |
| 730 | h32 ^= h32 >> 16; |
| 731 | |
| 732 | return h32; |
| 733 | } |
| 734 | |
| 735 | |
| 736 | U32 XXH32_digest (const XXH32_state_t* state_in) |
| 737 | { |
| 738 | XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; |
| 739 | |
| 740 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 741 | return XXH32_digest_endian(state_in, XXH_littleEndian); |
| 742 | else |
| 743 | return XXH32_digest_endian(state_in, XXH_bigEndian); |
| 744 | } |
| 745 | |
| 746 | |
| 747 | FORCE_INLINE XXH_errorcode XXH64_update_endian (XXH64_state_t* state_in, const void* input, size_t len, XXH_endianess endian) |
| 748 | { |
| 749 | XXH_istate64_t * state = (XXH_istate64_t *) state_in; |
Vincent Bernat | 3c2f2f2 | 2016-04-03 13:48:42 +0200 | [diff] [blame] | 750 | const BYTE* p = input; |
Willy Tarreau | b5684e0 | 2015-04-27 11:59:40 +0200 | [diff] [blame] | 751 | const BYTE* const bEnd = p + len; |
| 752 | |
| 753 | #ifdef XXH_ACCEPT_NULL_INPUT_POINTER |
| 754 | if (input==NULL) return XXH_ERROR; |
| 755 | #endif |
| 756 | |
| 757 | state->total_len += len; |
| 758 | |
| 759 | if (state->memsize + len < 32) // fill in tmp buffer |
| 760 | { |
| 761 | XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len); |
| 762 | state->memsize += (U32)len; |
| 763 | return XXH_OK; |
| 764 | } |
| 765 | |
| 766 | if (state->memsize) // some data left from previous update |
| 767 | { |
| 768 | XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize); |
| 769 | { |
| 770 | const U64* p64 = state->mem64; |
| 771 | state->v1 += XXH_readLE64(p64, endian) * PRIME64_2; |
| 772 | state->v1 = XXH_rotl64(state->v1, 31); |
| 773 | state->v1 *= PRIME64_1; |
| 774 | p64++; |
| 775 | state->v2 += XXH_readLE64(p64, endian) * PRIME64_2; |
| 776 | state->v2 = XXH_rotl64(state->v2, 31); |
| 777 | state->v2 *= PRIME64_1; |
| 778 | p64++; |
| 779 | state->v3 += XXH_readLE64(p64, endian) * PRIME64_2; |
| 780 | state->v3 = XXH_rotl64(state->v3, 31); |
| 781 | state->v3 *= PRIME64_1; |
| 782 | p64++; |
| 783 | state->v4 += XXH_readLE64(p64, endian) * PRIME64_2; |
| 784 | state->v4 = XXH_rotl64(state->v4, 31); |
| 785 | state->v4 *= PRIME64_1; |
| 786 | p64++; |
| 787 | } |
| 788 | p += 32-state->memsize; |
| 789 | state->memsize = 0; |
| 790 | } |
| 791 | |
| 792 | if (p+32 <= bEnd) |
| 793 | { |
| 794 | const BYTE* const limit = bEnd - 32; |
| 795 | U64 v1 = state->v1; |
| 796 | U64 v2 = state->v2; |
| 797 | U64 v3 = state->v3; |
| 798 | U64 v4 = state->v4; |
| 799 | |
| 800 | do |
| 801 | { |
| 802 | v1 += XXH_readLE64(p, endian) * PRIME64_2; |
| 803 | v1 = XXH_rotl64(v1, 31); |
| 804 | v1 *= PRIME64_1; |
| 805 | p+=8; |
| 806 | v2 += XXH_readLE64(p, endian) * PRIME64_2; |
| 807 | v2 = XXH_rotl64(v2, 31); |
| 808 | v2 *= PRIME64_1; |
| 809 | p+=8; |
| 810 | v3 += XXH_readLE64(p, endian) * PRIME64_2; |
| 811 | v3 = XXH_rotl64(v3, 31); |
| 812 | v3 *= PRIME64_1; |
| 813 | p+=8; |
| 814 | v4 += XXH_readLE64(p, endian) * PRIME64_2; |
| 815 | v4 = XXH_rotl64(v4, 31); |
| 816 | v4 *= PRIME64_1; |
| 817 | p+=8; |
| 818 | } |
| 819 | while (p<=limit); |
| 820 | |
| 821 | state->v1 = v1; |
| 822 | state->v2 = v2; |
| 823 | state->v3 = v3; |
| 824 | state->v4 = v4; |
| 825 | } |
| 826 | |
| 827 | if (p < bEnd) |
| 828 | { |
| 829 | XXH_memcpy(state->mem64, p, bEnd-p); |
| 830 | state->memsize = (int)(bEnd-p); |
| 831 | } |
| 832 | |
| 833 | return XXH_OK; |
| 834 | } |
| 835 | |
| 836 | XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void* input, size_t len) |
| 837 | { |
| 838 | XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; |
| 839 | |
| 840 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 841 | return XXH64_update_endian(state_in, input, len, XXH_littleEndian); |
| 842 | else |
| 843 | return XXH64_update_endian(state_in, input, len, XXH_bigEndian); |
| 844 | } |
| 845 | |
| 846 | |
| 847 | |
| 848 | FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state_in, XXH_endianess endian) |
| 849 | { |
| 850 | XXH_istate64_t * state = (XXH_istate64_t *) state_in; |
| 851 | const BYTE * p = (const BYTE*)state->mem64; |
| 852 | BYTE* bEnd = (BYTE*)state->mem64 + state->memsize; |
| 853 | U64 h64; |
| 854 | |
| 855 | if (state->total_len >= 32) |
| 856 | { |
| 857 | U64 v1 = state->v1; |
| 858 | U64 v2 = state->v2; |
| 859 | U64 v3 = state->v3; |
| 860 | U64 v4 = state->v4; |
| 861 | |
| 862 | h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18); |
| 863 | |
| 864 | v1 *= PRIME64_2; |
| 865 | v1 = XXH_rotl64(v1, 31); |
| 866 | v1 *= PRIME64_1; |
| 867 | h64 ^= v1; |
| 868 | h64 = h64*PRIME64_1 + PRIME64_4; |
| 869 | |
| 870 | v2 *= PRIME64_2; |
| 871 | v2 = XXH_rotl64(v2, 31); |
| 872 | v2 *= PRIME64_1; |
| 873 | h64 ^= v2; |
| 874 | h64 = h64*PRIME64_1 + PRIME64_4; |
| 875 | |
| 876 | v3 *= PRIME64_2; |
| 877 | v3 = XXH_rotl64(v3, 31); |
| 878 | v3 *= PRIME64_1; |
| 879 | h64 ^= v3; |
| 880 | h64 = h64*PRIME64_1 + PRIME64_4; |
| 881 | |
| 882 | v4 *= PRIME64_2; |
| 883 | v4 = XXH_rotl64(v4, 31); |
| 884 | v4 *= PRIME64_1; |
| 885 | h64 ^= v4; |
| 886 | h64 = h64*PRIME64_1 + PRIME64_4; |
| 887 | } |
| 888 | else |
| 889 | { |
| 890 | h64 = state->seed + PRIME64_5; |
| 891 | } |
| 892 | |
| 893 | h64 += (U64) state->total_len; |
| 894 | |
| 895 | while (p+8<=bEnd) |
| 896 | { |
| 897 | U64 k1 = XXH_readLE64(p, endian); |
| 898 | k1 *= PRIME64_2; |
| 899 | k1 = XXH_rotl64(k1,31); |
| 900 | k1 *= PRIME64_1; |
| 901 | h64 ^= k1; |
| 902 | h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; |
| 903 | p+=8; |
| 904 | } |
| 905 | |
| 906 | if (p+4<=bEnd) |
| 907 | { |
| 908 | h64 ^= (U64)(XXH_readLE32(p, endian)) * PRIME64_1; |
| 909 | h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; |
| 910 | p+=4; |
| 911 | } |
| 912 | |
| 913 | while (p<bEnd) |
| 914 | { |
| 915 | h64 ^= (*p) * PRIME64_5; |
| 916 | h64 = XXH_rotl64(h64, 11) * PRIME64_1; |
| 917 | p++; |
| 918 | } |
| 919 | |
| 920 | h64 ^= h64 >> 33; |
| 921 | h64 *= PRIME64_2; |
| 922 | h64 ^= h64 >> 29; |
| 923 | h64 *= PRIME64_3; |
| 924 | h64 ^= h64 >> 32; |
| 925 | |
| 926 | return h64; |
| 927 | } |
| 928 | |
| 929 | |
| 930 | unsigned long long XXH64_digest (const XXH64_state_t* state_in) |
| 931 | { |
| 932 | XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; |
| 933 | |
| 934 | if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) |
| 935 | return XXH64_digest_endian(state_in, XXH_littleEndian); |
| 936 | else |
| 937 | return XXH64_digest_endian(state_in, XXH_bigEndian); |
| 938 | } |
| 939 | |
| 940 | |