1/* 2 * xxHash - Fast Hash algorithm 3 * Copyright (c) 2012-2020, Yann Collet, Facebook, Inc. 4 * 5 * You can contact the author at : 6 * - xxHash homepage: http://www.xxhash.com 7 * - xxHash source repository : https://github.com/Cyan4973/xxHash 8 * 9 * This source code is licensed under both the BSD-style license (found in the 10 * LICENSE file in the root directory of this source tree) and the GPLv2 (found 11 * in the COPYING file in the root directory of this source tree). 12 * You may select, at your option, one of the above-listed licenses. 13*/ 14 15 16/* ************************************* 17* Tuning parameters 18***************************************/ 19/*!XXH_FORCE_MEMORY_ACCESS : 20 * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable. 21 * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal. 22 * The below switch allow to select different access method for improved performance. 23 * Method 0 (default) : use `memcpy()`. Safe and portable. 24 * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable). 25 * This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`. 26 * Method 2 : direct access. This method doesn't depend on compiler but violate C standard. 27 * It can generate buggy code on targets which do not support unaligned memory accesses. 28 * But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6) 29 * See http://stackoverflow.com/a/32095106/646947 for details. 30 * Prefer these methods in priority order (0 > 1 > 2) 31 */ 32#ifndef XXH_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */ 33# if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) ) 34# define XXH_FORCE_MEMORY_ACCESS 2 35# elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \ 36 (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) )) || \ 37 defined(__ICCARM__) 38# define XXH_FORCE_MEMORY_ACCESS 1 39# endif 40#endif 41 42/*!XXH_ACCEPT_NULL_INPUT_POINTER : 43 * If the input pointer is a null pointer, xxHash default behavior is to trigger a memory access error, since it is a bad pointer. 44 * When this option is enabled, xxHash output for null input pointers will be the same as a null-length input. 45 * By default, this option is disabled. To enable it, uncomment below define : 46 */ 47/* #define XXH_ACCEPT_NULL_INPUT_POINTER 1 */ 48 49/*!XXH_FORCE_NATIVE_FORMAT : 50 * By default, xxHash library provides endian-independent Hash values, based on little-endian convention. 51 * Results are therefore identical for little-endian and big-endian CPU. 52 * This comes at a performance cost for big-endian CPU, since some swapping is required to emulate little-endian format. 53 * Should endian-independence be of no importance for your application, you may set the #define below to 1, 54 * to improve speed for Big-endian CPU. 55 * This option has no impact on Little_Endian CPU. 56 */ 57#ifndef XXH_FORCE_NATIVE_FORMAT /* can be defined externally */ 58# define XXH_FORCE_NATIVE_FORMAT 0 59#endif 60 61/*!XXH_FORCE_ALIGN_CHECK : 62 * This is a minor performance trick, only useful with lots of very small keys. 63 * It means : check for aligned/unaligned input. 64 * The check costs one initial branch per hash; set to 0 when the input data 65 * is guaranteed to be aligned. 66 */ 67#ifndef XXH_FORCE_ALIGN_CHECK /* can be defined externally */ 68# if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64) 69# define XXH_FORCE_ALIGN_CHECK 0 70# else 71# define XXH_FORCE_ALIGN_CHECK 1 72# endif 73#endif 74 75 76/* ************************************* 77* Includes & Memory related functions 78***************************************/ 79/* Modify the local functions below should you wish to use some other memory routines */ 80/* for malloc(), free() */ 81#include <stdlib.h> 82#include <stddef.h> /* size_t */ 83static void* XXH_malloc(size_t s) { return malloc(s); } 84static void XXH_free (void* p) { free(p); } 85/* for memcpy() */ 86#include <string.h> 87static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); } 88 89#ifndef XXH_STATIC_LINKING_ONLY 90# define XXH_STATIC_LINKING_ONLY 91#endif 92#include "xxhash.h" 93 94 95/* ************************************* 96* Compiler Specific Options 97***************************************/ 98#if (defined(__GNUC__) && !defined(__STRICT_ANSI__)) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */ 99# define INLINE_KEYWORD inline 100#else 101# define INLINE_KEYWORD 102#endif 103 104#if defined(__GNUC__) || defined(__ICCARM__) 105# define FORCE_INLINE_ATTR __attribute__((always_inline)) 106#elif defined(_MSC_VER) 107# define FORCE_INLINE_ATTR __forceinline 108#else 109# define FORCE_INLINE_ATTR 110#endif 111 112#define FORCE_INLINE_TEMPLATE static INLINE_KEYWORD FORCE_INLINE_ATTR 113 114 115#ifdef _MSC_VER 116# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */ 117#endif 118 119 120/* ************************************* 121* Basic Types 122***************************************/ 123#ifndef MEM_MODULE 124# define MEM_MODULE 125# if !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) ) 126# include <stdint.h> 127 typedef uint8_t BYTE; 128 typedef uint16_t U16; 129 typedef uint32_t U32; 130 typedef int32_t S32; 131 typedef uint64_t U64; 132# else 133 typedef unsigned char BYTE; 134 typedef unsigned short U16; 135 typedef unsigned int U32; 136 typedef signed int S32; 137 typedef unsigned long long U64; /* if your compiler doesn't support unsigned long long, replace by another 64-bit type here. Note that xxhash.h will also need to be updated. */ 138# endif 139#endif 140 141 142#if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2)) 143 144/* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */ 145static U32 XXH_read32(const void* memPtr) { return *(const U32*) memPtr; } 146static U64 XXH_read64(const void* memPtr) { return *(const U64*) memPtr; } 147 148#elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1)) 149 150/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */ 151/* currently only defined for gcc and icc */ 152typedef union { U32 u32; U64 u64; } __attribute__((packed)) unalign; 153 154static U32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; } 155static U64 XXH_read64(const void* ptr) { return ((const unalign*)ptr)->u64; } 156 157#else 158 159/* portable and safe solution. Generally efficient. 160 * see : http://stackoverflow.com/a/32095106/646947 161 */ 162 163static U32 XXH_read32(const void* memPtr) 164{ 165 U32 val; 166 memcpy(&val, memPtr, sizeof(val)); 167 return val; 168} 169 170static U64 XXH_read64(const void* memPtr) 171{ 172 U64 val; 173 memcpy(&val, memPtr, sizeof(val)); 174 return val; 175} 176 177#endif /* XXH_FORCE_DIRECT_MEMORY_ACCESS */ 178 179 180/* **************************************** 181* Compiler-specific Functions and Macros 182******************************************/ 183#define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__) 184 185/* Note : although _rotl exists for minGW (GCC under windows), performance seems poor */ 186#if defined(_MSC_VER) 187# define XXH_rotl32(x,r) _rotl(x,r) 188# define XXH_rotl64(x,r) _rotl64(x,r) 189#else 190#if defined(__ICCARM__) 191# include <intrinsics.h> 192# define XXH_rotl32(x,r) __ROR(x,(32 - r)) 193#else 194# define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r))) 195#endif 196# define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r))) 197#endif 198 199#if defined(_MSC_VER) /* Visual Studio */ 200# define XXH_swap32 _byteswap_ulong 201# define XXH_swap64 _byteswap_uint64 202#elif GCC_VERSION >= 403 203# define XXH_swap32 __builtin_bswap32 204# define XXH_swap64 __builtin_bswap64 205#else 206static U32 XXH_swap32 (U32 x) 207{ 208 return ((x << 24) & 0xff000000 ) | 209 ((x << 8) & 0x00ff0000 ) | 210 ((x >> 8) & 0x0000ff00 ) | 211 ((x >> 24) & 0x000000ff ); 212} 213static U64 XXH_swap64 (U64 x) 214{ 215 return ((x << 56) & 0xff00000000000000ULL) | 216 ((x << 40) & 0x00ff000000000000ULL) | 217 ((x << 24) & 0x0000ff0000000000ULL) | 218 ((x << 8) & 0x000000ff00000000ULL) | 219 ((x >> 8) & 0x00000000ff000000ULL) | 220 ((x >> 24) & 0x0000000000ff0000ULL) | 221 ((x >> 40) & 0x000000000000ff00ULL) | 222 ((x >> 56) & 0x00000000000000ffULL); 223} 224#endif 225 226 227/* ************************************* 228* Architecture Macros 229***************************************/ 230typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess; 231 232/* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */ 233#ifndef XXH_CPU_LITTLE_ENDIAN 234 static const int g_one = 1; 235# define XXH_CPU_LITTLE_ENDIAN (*(const char*)(&g_one)) 236#endif 237 238 239/* *************************** 240* Memory reads 241*****************************/ 242typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment; 243 244FORCE_INLINE_TEMPLATE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align) 245{ 246 if (align==XXH_unaligned) 247 return endian==XXH_littleEndian ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr)); 248 else 249 return endian==XXH_littleEndian ? *(const U32*)ptr : XXH_swap32(*(const U32*)ptr); 250} 251 252FORCE_INLINE_TEMPLATE U32 XXH_readLE32(const void* ptr, XXH_endianess endian) 253{ 254 return XXH_readLE32_align(ptr, endian, XXH_unaligned); 255} 256 257static U32 XXH_readBE32(const void* ptr) 258{ 259 return XXH_CPU_LITTLE_ENDIAN ? XXH_swap32(XXH_read32(ptr)) : XXH_read32(ptr); 260} 261 262FORCE_INLINE_TEMPLATE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align) 263{ 264 if (align==XXH_unaligned) 265 return endian==XXH_littleEndian ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr)); 266 else 267 return endian==XXH_littleEndian ? *(const U64*)ptr : XXH_swap64(*(const U64*)ptr); 268} 269 270FORCE_INLINE_TEMPLATE U64 XXH_readLE64(const void* ptr, XXH_endianess endian) 271{ 272 return XXH_readLE64_align(ptr, endian, XXH_unaligned); 273} 274 275static U64 XXH_readBE64(const void* ptr) 276{ 277 return XXH_CPU_LITTLE_ENDIAN ? XXH_swap64(XXH_read64(ptr)) : XXH_read64(ptr); 278} 279 280 281/* ************************************* 282* Macros 283***************************************/ 284#define XXH_STATIC_ASSERT(c) { enum { XXH_static_assert = 1/(int)(!!(c)) }; } /* use only *after* variable declarations */ 285 286 287/* ************************************* 288* Constants 289***************************************/ 290static const U32 PRIME32_1 = 2654435761U; 291static const U32 PRIME32_2 = 2246822519U; 292static const U32 PRIME32_3 = 3266489917U; 293static const U32 PRIME32_4 = 668265263U; 294static const U32 PRIME32_5 = 374761393U; 295 296static const U64 PRIME64_1 = 11400714785074694791ULL; 297static const U64 PRIME64_2 = 14029467366897019727ULL; 298static const U64 PRIME64_3 = 1609587929392839161ULL; 299static const U64 PRIME64_4 = 9650029242287828579ULL; 300static const U64 PRIME64_5 = 2870177450012600261ULL; 301 302XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; } 303 304 305/* ************************** 306* Utils 307****************************/ 308XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* restrict dstState, const XXH32_state_t* restrict srcState) 309{ 310 memcpy(dstState, srcState, sizeof(*dstState)); 311} 312 313XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* restrict dstState, const XXH64_state_t* restrict srcState) 314{ 315 memcpy(dstState, srcState, sizeof(*dstState)); 316} 317 318 319/* *************************** 320* Simple Hash Functions 321*****************************/ 322 323static U32 XXH32_round(U32 seed, U32 input) 324{ 325 seed += input * PRIME32_2; 326 seed = XXH_rotl32(seed, 13); 327 seed *= PRIME32_1; 328 return seed; 329} 330 331FORCE_INLINE_TEMPLATE U32 XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_endianess endian, XXH_alignment align) 332{ 333 const BYTE* p = (const BYTE*)input; 334 const BYTE* bEnd = p + len; 335 U32 h32; 336#define XXH_get32bits(p) XXH_readLE32_align(p, endian, align) 337 338#ifdef XXH_ACCEPT_NULL_INPUT_POINTER 339 if (p==NULL) { 340 len=0; 341 bEnd=p=(const BYTE*)(size_t)16; 342 } 343#endif 344 345 if (len>=16) { 346 const BYTE* const limit = bEnd - 16; 347 U32 v1 = seed + PRIME32_1 + PRIME32_2; 348 U32 v2 = seed + PRIME32_2; 349 U32 v3 = seed + 0; 350 U32 v4 = seed - PRIME32_1; 351 352 do { 353 v1 = XXH32_round(v1, XXH_get32bits(p)); p+=4; 354 v2 = XXH32_round(v2, XXH_get32bits(p)); p+=4; 355 v3 = XXH32_round(v3, XXH_get32bits(p)); p+=4; 356 v4 = XXH32_round(v4, XXH_get32bits(p)); p+=4; 357 } while (p<=limit); 358 359 h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18); 360 } else { 361 h32 = seed + PRIME32_5; 362 } 363 364 h32 += (U32) len; 365 366 while (p+4<=bEnd) { 367 h32 += XXH_get32bits(p) * PRIME32_3; 368 h32 = XXH_rotl32(h32, 17) * PRIME32_4 ; 369 p+=4; 370 } 371 372 while (p<bEnd) { 373 h32 += (*p) * PRIME32_5; 374 h32 = XXH_rotl32(h32, 11) * PRIME32_1 ; 375 p++; 376 } 377 378 h32 ^= h32 >> 15; 379 h32 *= PRIME32_2; 380 h32 ^= h32 >> 13; 381 h32 *= PRIME32_3; 382 h32 ^= h32 >> 16; 383 384 return h32; 385} 386 387 388XXH_PUBLIC_API unsigned int XXH32 (const void* input, size_t len, unsigned int seed) 389{ 390#if 0 391 /* Simple version, good for code maintenance, but unfortunately slow for small inputs */ 392 XXH32_CREATESTATE_STATIC(state); 393 XXH32_reset(state, seed); 394 XXH32_update(state, input, len); 395 return XXH32_digest(state); 396#else 397 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; 398 399 if (XXH_FORCE_ALIGN_CHECK) { 400 if ((((size_t)input) & 3) == 0) { /* Input is 4-bytes aligned, leverage the speed benefit */ 401 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 402 return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned); 403 else 404 return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned); 405 } } 406 407 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 408 return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned); 409 else 410 return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned); 411#endif 412} 413 414 415static U64 XXH64_round(U64 acc, U64 input) 416{ 417 acc += input * PRIME64_2; 418 acc = XXH_rotl64(acc, 31); 419 acc *= PRIME64_1; 420 return acc; 421} 422 423static U64 XXH64_mergeRound(U64 acc, U64 val) 424{ 425 val = XXH64_round(0, val); 426 acc ^= val; 427 acc = acc * PRIME64_1 + PRIME64_4; 428 return acc; 429} 430 431FORCE_INLINE_TEMPLATE U64 XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_endianess endian, XXH_alignment align) 432{ 433 const BYTE* p = (const BYTE*)input; 434 const BYTE* const bEnd = p + len; 435 U64 h64; 436#define XXH_get64bits(p) XXH_readLE64_align(p, endian, align) 437 438#ifdef XXH_ACCEPT_NULL_INPUT_POINTER 439 if (p==NULL) { 440 len=0; 441 bEnd=p=(const BYTE*)(size_t)32; 442 } 443#endif 444 445 if (len>=32) { 446 const BYTE* const limit = bEnd - 32; 447 U64 v1 = seed + PRIME64_1 + PRIME64_2; 448 U64 v2 = seed + PRIME64_2; 449 U64 v3 = seed + 0; 450 U64 v4 = seed - PRIME64_1; 451 452 do { 453 v1 = XXH64_round(v1, XXH_get64bits(p)); p+=8; 454 v2 = XXH64_round(v2, XXH_get64bits(p)); p+=8; 455 v3 = XXH64_round(v3, XXH_get64bits(p)); p+=8; 456 v4 = XXH64_round(v4, XXH_get64bits(p)); p+=8; 457 } while (p<=limit); 458 459 h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18); 460 h64 = XXH64_mergeRound(h64, v1); 461 h64 = XXH64_mergeRound(h64, v2); 462 h64 = XXH64_mergeRound(h64, v3); 463 h64 = XXH64_mergeRound(h64, v4); 464 465 } else { 466 h64 = seed + PRIME64_5; 467 } 468 469 h64 += (U64) len; 470 471 while (p+8<=bEnd) { 472 U64 const k1 = XXH64_round(0, XXH_get64bits(p)); 473 h64 ^= k1; 474 h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; 475 p+=8; 476 } 477 478 if (p+4<=bEnd) { 479 h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1; 480 h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; 481 p+=4; 482 } 483 484 while (p<bEnd) { 485 h64 ^= (*p) * PRIME64_5; 486 h64 = XXH_rotl64(h64, 11) * PRIME64_1; 487 p++; 488 } 489 490 h64 ^= h64 >> 33; 491 h64 *= PRIME64_2; 492 h64 ^= h64 >> 29; 493 h64 *= PRIME64_3; 494 h64 ^= h64 >> 32; 495 496 return h64; 497} 498 499 500XXH_PUBLIC_API unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed) 501{ 502#if 0 503 /* Simple version, good for code maintenance, but unfortunately slow for small inputs */ 504 XXH64_CREATESTATE_STATIC(state); 505 XXH64_reset(state, seed); 506 XXH64_update(state, input, len); 507 return XXH64_digest(state); 508#else 509 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; 510 511 if (XXH_FORCE_ALIGN_CHECK) { 512 if ((((size_t)input) & 7)==0) { /* Input is aligned, let's leverage the speed advantage */ 513 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 514 return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned); 515 else 516 return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned); 517 } } 518 519 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 520 return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned); 521 else 522 return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned); 523#endif 524} 525 526 527/* ************************************************** 528* Advanced Hash Functions 529****************************************************/ 530 531XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void) 532{ 533 return (XXH32_state_t*)XXH_malloc(sizeof(XXH32_state_t)); 534} 535XXH_PUBLIC_API XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr) 536{ 537 XXH_free(statePtr); 538 return XXH_OK; 539} 540 541XXH_PUBLIC_API XXH64_state_t* XXH64_createState(void) 542{ 543 return (XXH64_state_t*)XXH_malloc(sizeof(XXH64_state_t)); 544} 545XXH_PUBLIC_API XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr) 546{ 547 XXH_free(statePtr); 548 return XXH_OK; 549} 550 551 552/*** Hash feed ***/ 553 554XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, unsigned int seed) 555{ 556 XXH32_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */ 557 memset(&state, 0, sizeof(state)-4); /* do not write into reserved, for future removal */ 558 state.v1 = seed + PRIME32_1 + PRIME32_2; 559 state.v2 = seed + PRIME32_2; 560 state.v3 = seed + 0; 561 state.v4 = seed - PRIME32_1; 562 memcpy(statePtr, &state, sizeof(state)); 563 return XXH_OK; 564} 565 566 567XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long long seed) 568{ 569 XXH64_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */ 570 memset(&state, 0, sizeof(state)-8); /* do not write into reserved, for future removal */ 571 state.v1 = seed + PRIME64_1 + PRIME64_2; 572 state.v2 = seed + PRIME64_2; 573 state.v3 = seed + 0; 574 state.v4 = seed - PRIME64_1; 575 memcpy(statePtr, &state, sizeof(state)); 576 return XXH_OK; 577} 578 579 580FORCE_INLINE_TEMPLATE XXH_errorcode XXH32_update_endian (XXH32_state_t* state, const void* input, size_t len, XXH_endianess endian) 581{ 582 const BYTE* p = (const BYTE*)input; 583 const BYTE* const bEnd = p + len; 584 585#ifdef XXH_ACCEPT_NULL_INPUT_POINTER 586 if (input==NULL) return XXH_ERROR; 587#endif 588 589 state->total_len_32 += (unsigned)len; 590 state->large_len |= (len>=16) | (state->total_len_32>=16); 591 592 if (state->memsize + len < 16) { /* fill in tmp buffer */ 593 XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len); 594 state->memsize += (unsigned)len; 595 return XXH_OK; 596 } 597 598 if (state->memsize) { /* some data left from previous update */ 599 XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize); 600 { const U32* p32 = state->mem32; 601 state->v1 = XXH32_round(state->v1, XXH_readLE32(p32, endian)); p32++; 602 state->v2 = XXH32_round(state->v2, XXH_readLE32(p32, endian)); p32++; 603 state->v3 = XXH32_round(state->v3, XXH_readLE32(p32, endian)); p32++; 604 state->v4 = XXH32_round(state->v4, XXH_readLE32(p32, endian)); p32++; 605 } 606 p += 16-state->memsize; 607 state->memsize = 0; 608 } 609 610 if (p <= bEnd-16) { 611 const BYTE* const limit = bEnd - 16; 612 U32 v1 = state->v1; 613 U32 v2 = state->v2; 614 U32 v3 = state->v3; 615 U32 v4 = state->v4; 616 617 do { 618 v1 = XXH32_round(v1, XXH_readLE32(p, endian)); p+=4; 619 v2 = XXH32_round(v2, XXH_readLE32(p, endian)); p+=4; 620 v3 = XXH32_round(v3, XXH_readLE32(p, endian)); p+=4; 621 v4 = XXH32_round(v4, XXH_readLE32(p, endian)); p+=4; 622 } while (p<=limit); 623 624 state->v1 = v1; 625 state->v2 = v2; 626 state->v3 = v3; 627 state->v4 = v4; 628 } 629 630 if (p < bEnd) { 631 XXH_memcpy(state->mem32, p, (size_t)(bEnd-p)); 632 state->memsize = (unsigned)(bEnd-p); 633 } 634 635 return XXH_OK; 636} 637 638XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len) 639{ 640 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; 641 642 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 643 return XXH32_update_endian(state_in, input, len, XXH_littleEndian); 644 else 645 return XXH32_update_endian(state_in, input, len, XXH_bigEndian); 646} 647 648 649 650FORCE_INLINE_TEMPLATE U32 XXH32_digest_endian (const XXH32_state_t* state, XXH_endianess endian) 651{ 652 const BYTE * p = (const BYTE*)state->mem32; 653 const BYTE* const bEnd = (const BYTE*)(state->mem32) + state->memsize; 654 U32 h32; 655 656 if (state->large_len) { 657 h32 = XXH_rotl32(state->v1, 1) + XXH_rotl32(state->v2, 7) + XXH_rotl32(state->v3, 12) + XXH_rotl32(state->v4, 18); 658 } else { 659 h32 = state->v3 /* == seed */ + PRIME32_5; 660 } 661 662 h32 += state->total_len_32; 663 664 while (p+4<=bEnd) { 665 h32 += XXH_readLE32(p, endian) * PRIME32_3; 666 h32 = XXH_rotl32(h32, 17) * PRIME32_4; 667 p+=4; 668 } 669 670 while (p<bEnd) { 671 h32 += (*p) * PRIME32_5; 672 h32 = XXH_rotl32(h32, 11) * PRIME32_1; 673 p++; 674 } 675 676 h32 ^= h32 >> 15; 677 h32 *= PRIME32_2; 678 h32 ^= h32 >> 13; 679 h32 *= PRIME32_3; 680 h32 ^= h32 >> 16; 681 682 return h32; 683} 684 685 686XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state_in) 687{ 688 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; 689 690 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 691 return XXH32_digest_endian(state_in, XXH_littleEndian); 692 else 693 return XXH32_digest_endian(state_in, XXH_bigEndian); 694} 695 696 697 698/* **** XXH64 **** */ 699 700FORCE_INLINE_TEMPLATE XXH_errorcode XXH64_update_endian (XXH64_state_t* state, const void* input, size_t len, XXH_endianess endian) 701{ 702 const BYTE* p = (const BYTE*)input; 703 const BYTE* const bEnd = p + len; 704 705#ifdef XXH_ACCEPT_NULL_INPUT_POINTER 706 if (input==NULL) return XXH_ERROR; 707#endif 708 709 state->total_len += len; 710 711 if (state->memsize + len < 32) { /* fill in tmp buffer */ 712 if (input != NULL) { 713 XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len); 714 } 715 state->memsize += (U32)len; 716 return XXH_OK; 717 } 718 719 if (state->memsize) { /* tmp buffer is full */ 720 XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize); 721 state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0, endian)); 722 state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1, endian)); 723 state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2, endian)); 724 state->v4 = XXH64_round(state->v4, XXH_readLE64(state->mem64+3, endian)); 725 p += 32-state->memsize; 726 state->memsize = 0; 727 } 728 729 if (p+32 <= bEnd) { 730 const BYTE* const limit = bEnd - 32; 731 U64 v1 = state->v1; 732 U64 v2 = state->v2; 733 U64 v3 = state->v3; 734 U64 v4 = state->v4; 735 736 do { 737 v1 = XXH64_round(v1, XXH_readLE64(p, endian)); p+=8; 738 v2 = XXH64_round(v2, XXH_readLE64(p, endian)); p+=8; 739 v3 = XXH64_round(v3, XXH_readLE64(p, endian)); p+=8; 740 v4 = XXH64_round(v4, XXH_readLE64(p, endian)); p+=8; 741 } while (p<=limit); 742 743 state->v1 = v1; 744 state->v2 = v2; 745 state->v3 = v3; 746 state->v4 = v4; 747 } 748 749 if (p < bEnd) { 750 XXH_memcpy(state->mem64, p, (size_t)(bEnd-p)); 751 state->memsize = (unsigned)(bEnd-p); 752 } 753 754 return XXH_OK; 755} 756 757XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void* input, size_t len) 758{ 759 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; 760 761 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 762 return XXH64_update_endian(state_in, input, len, XXH_littleEndian); 763 else 764 return XXH64_update_endian(state_in, input, len, XXH_bigEndian); 765} 766 767 768 769FORCE_INLINE_TEMPLATE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endianess endian) 770{ 771 const BYTE * p = (const BYTE*)state->mem64; 772 const BYTE* const bEnd = (const BYTE*)state->mem64 + state->memsize; 773 U64 h64; 774 775 if (state->total_len >= 32) { 776 U64 const v1 = state->v1; 777 U64 const v2 = state->v2; 778 U64 const v3 = state->v3; 779 U64 const v4 = state->v4; 780 781 h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18); 782 h64 = XXH64_mergeRound(h64, v1); 783 h64 = XXH64_mergeRound(h64, v2); 784 h64 = XXH64_mergeRound(h64, v3); 785 h64 = XXH64_mergeRound(h64, v4); 786 } else { 787 h64 = state->v3 + PRIME64_5; 788 } 789 790 h64 += (U64) state->total_len; 791 792 while (p+8<=bEnd) { 793 U64 const k1 = XXH64_round(0, XXH_readLE64(p, endian)); 794 h64 ^= k1; 795 h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; 796 p+=8; 797 } 798 799 if (p+4<=bEnd) { 800 h64 ^= (U64)(XXH_readLE32(p, endian)) * PRIME64_1; 801 h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3; 802 p+=4; 803 } 804 805 while (p<bEnd) { 806 h64 ^= (*p) * PRIME64_5; 807 h64 = XXH_rotl64(h64, 11) * PRIME64_1; 808 p++; 809 } 810 811 h64 ^= h64 >> 33; 812 h64 *= PRIME64_2; 813 h64 ^= h64 >> 29; 814 h64 *= PRIME64_3; 815 h64 ^= h64 >> 32; 816 817 return h64; 818} 819 820 821XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state_in) 822{ 823 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN; 824 825 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT) 826 return XXH64_digest_endian(state_in, XXH_littleEndian); 827 else 828 return XXH64_digest_endian(state_in, XXH_bigEndian); 829} 830 831 832/* ************************** 833* Canonical representation 834****************************/ 835 836/*! Default XXH result types are basic unsigned 32 and 64 bits. 837* The canonical representation follows human-readable write convention, aka big-endian (large digits first). 838* These functions allow transformation of hash result into and from its canonical format. 839* This way, hash values can be written into a file or buffer, and remain comparable across different systems and programs. 840*/ 841 842XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash) 843{ 844 XXH_STATIC_ASSERT(sizeof(XXH32_canonical_t) == sizeof(XXH32_hash_t)); 845 if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap32(hash); 846 memcpy(dst, &hash, sizeof(*dst)); 847} 848 849XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t hash) 850{ 851 XXH_STATIC_ASSERT(sizeof(XXH64_canonical_t) == sizeof(XXH64_hash_t)); 852 if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap64(hash); 853 memcpy(dst, &hash, sizeof(*dst)); 854} 855 856XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src) 857{ 858 return XXH_readBE32(src); 859} 860 861XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src) 862{ 863 return XXH_readBE64(src); 864} 865