1 /*
2 * xxHash - Fast Hash algorithm
3 * Copyright (C) 2012-2016, Yann Collet
4 *
5 * BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions are
9 * met:
10 *
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above
14 * copyright notice, this list of conditions and the following disclaimer
15 * in the documentation and/or other materials provided with the
16 * distribution.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 *
30 * You can contact the author at :
31 * - xxHash homepage: http://www.xxhash.com
32 * - xxHash source repository : https://github.com/Cyan4973/xxHash
33 */
34
35
36 /* *************************************
37 * Tuning parameters
38 ***************************************/
39 /*!XXH_FORCE_MEMORY_ACCESS :
40 * By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
41 * Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
42 * The below switch allow to select different access method for improved performance.
43 * Method 0 (default) : use `memcpy()`. Safe and portable.
44 * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
45 * This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
46 * Method 2 : direct access. This method doesn't depend on compiler but violate C standard.
47 * It can generate buggy code on targets which do not support unaligned memory accesses.
48 * But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
49 * See http://stackoverflow.com/a/32095106/646947 for details.
50 * Prefer these methods in priority order (0 > 1 > 2)
51 */
52 #ifndef XXH_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
53 # 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__) )
54 # define XXH_FORCE_MEMORY_ACCESS 2
55 # elif (defined(__INTEL_COMPILER) && !defined(WIN32)) || \
56 (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) )) || \
57 defined(__ICCARM__)
58 # define XXH_FORCE_MEMORY_ACCESS 1
59 # endif
60 #endif
61
62 /*!XXH_ACCEPT_NULL_INPUT_POINTER :
63 * If the input pointer is a null pointer, xxHash default behavior is to trigger a memory access error, since it is a bad pointer.
64 * When this option is enabled, xxHash output for null input pointers will be the same as a null-length input.
65 * By default, this option is disabled. To enable it, uncomment below define :
66 */
67 /* #define XXH_ACCEPT_NULL_INPUT_POINTER 1 */
68
69 /*!XXH_FORCE_NATIVE_FORMAT :
70 * By default, xxHash library provides endian-independent Hash values, based on little-endian convention.
71 * Results are therefore identical for little-endian and big-endian CPU.
72 * This comes at a performance cost for big-endian CPU, since some swapping is required to emulate little-endian format.
73 * Should endian-independence be of no importance for your application, you may set the #define below to 1,
74 * to improve speed for Big-endian CPU.
75 * This option has no impact on Little_Endian CPU.
76 */
77 #ifndef XXH_FORCE_NATIVE_FORMAT /* can be defined externally */
78 # define XXH_FORCE_NATIVE_FORMAT 0
79 #endif
80
81 /*!XXH_FORCE_ALIGN_CHECK :
82 * This is a minor performance trick, only useful with lots of very small keys.
83 * It means : check for aligned/unaligned input.
84 * The check costs one initial branch per hash; set to 0 when the input data
85 * is guaranteed to be aligned.
86 */
87 #ifndef XXH_FORCE_ALIGN_CHECK /* can be defined externally */
88 # if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64)
89 # define XXH_FORCE_ALIGN_CHECK 0
90 # else
91 # define XXH_FORCE_ALIGN_CHECK 1
92 # endif
93 #endif
94
95
96 /* *************************************
97 * Includes & Memory related functions
98 ***************************************/
99 /* Modify the local functions below should you wish to use some other memory routines */
100 /* for malloc(), free() */
101 #include <stdlib.h>
102 #include <stddef.h> /* size_t */
XXH_malloc(size_t s)103 static void* XXH_malloc(size_t s) { return malloc(s); }
XXH_free(void * p)104 static void XXH_free (void* p) { free(p); }
105 /* for memcpy() */
106 #include <string.h>
XXH_memcpy(void * dest,const void * src,size_t size)107 static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
108
109 #ifndef XXH_STATIC_LINKING_ONLY
110 # define XXH_STATIC_LINKING_ONLY
111 #endif
112 #include "xxhash.h"
113
114
115 /* *************************************
116 * Compiler Specific Options
117 ***************************************/
118 #if defined (__GNUC__) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */
119 # define INLINE_KEYWORD inline
120 #else
121 # define INLINE_KEYWORD
122 #endif
123
124 #if defined(__GNUC__) || defined(__ICCARM__)
125 # define FORCE_INLINE_ATTR __attribute__((always_inline))
126 #elif defined(_MSC_VER)
127 # define FORCE_INLINE_ATTR __forceinline
128 #else
129 # define FORCE_INLINE_ATTR
130 #endif
131
132 #define FORCE_INLINE_TEMPLATE static INLINE_KEYWORD FORCE_INLINE_ATTR
133
134
135 #ifdef _MSC_VER
136 # pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
137 #endif
138
139
140 /* *************************************
141 * Basic Types
142 ***************************************/
143 #ifndef MEM_MODULE
144 # define MEM_MODULE
145 # if !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
146 # include <stdint.h>
147 typedef uint8_t BYTE;
148 typedef uint16_t U16;
149 typedef uint32_t U32;
150 typedef int32_t S32;
151 typedef uint64_t U64;
152 # else
153 typedef unsigned char BYTE;
154 typedef unsigned short U16;
155 typedef unsigned int U32;
156 typedef signed int S32;
157 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. */
158 # endif
159 #endif
160
161
162 #if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2))
163
164 /* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */
XXH_read32(const void * memPtr)165 static U32 XXH_read32(const void* memPtr) { return *(const U32*) memPtr; }
XXH_read64(const void * memPtr)166 static U64 XXH_read64(const void* memPtr) { return *(const U64*) memPtr; }
167
168 #elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1))
169
170 /* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
171 /* currently only defined for gcc and icc */
172 typedef union { U32 u32; U64 u64; } __attribute__((packed)) unalign;
173
XXH_read32(const void * ptr)174 static U32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
XXH_read64(const void * ptr)175 static U64 XXH_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
176
177 #else
178
179 /* portable and safe solution. Generally efficient.
180 * see : http://stackoverflow.com/a/32095106/646947
181 */
182
XXH_read32(const void * memPtr)183 static U32 XXH_read32(const void* memPtr)
184 {
185 U32 val;
186 memcpy(&val, memPtr, sizeof(val));
187 return val;
188 }
189
XXH_read64(const void * memPtr)190 static U64 XXH_read64(const void* memPtr)
191 {
192 U64 val;
193 memcpy(&val, memPtr, sizeof(val));
194 return val;
195 }
196
197 #endif /* XXH_FORCE_DIRECT_MEMORY_ACCESS */
198
199
200 /* ****************************************
201 * Compiler-specific Functions and Macros
202 ******************************************/
203 #define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
204
205 /* Note : although _rotl exists for minGW (GCC under windows), performance seems poor */
206 #if defined(_MSC_VER)
207 # define XXH_rotl32(x,r) _rotl(x,r)
208 # define XXH_rotl64(x,r) _rotl64(x,r)
209 #else
210 #if defined(__ICCARM__)
211 # include <intrinsics.h>
212 # define XXH_rotl32(x,r) __ROR(x,(32 - r))
213 #else
214 # define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r)))
215 #endif
216 # define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r)))
217 #endif
218
219 #if defined(_MSC_VER) /* Visual Studio */
220 # define XXH_swap32 _byteswap_ulong
221 # define XXH_swap64 _byteswap_uint64
222 #elif GCC_VERSION >= 403
223 # define XXH_swap32 __builtin_bswap32
224 # define XXH_swap64 __builtin_bswap64
225 #else
XXH_swap32(U32 x)226 static U32 XXH_swap32 (U32 x)
227 {
228 return ((x << 24) & 0xff000000 ) |
229 ((x << 8) & 0x00ff0000 ) |
230 ((x >> 8) & 0x0000ff00 ) |
231 ((x >> 24) & 0x000000ff );
232 }
XXH_swap64(U64 x)233 static U64 XXH_swap64 (U64 x)
234 {
235 return ((x << 56) & 0xff00000000000000ULL) |
236 ((x << 40) & 0x00ff000000000000ULL) |
237 ((x << 24) & 0x0000ff0000000000ULL) |
238 ((x << 8) & 0x000000ff00000000ULL) |
239 ((x >> 8) & 0x00000000ff000000ULL) |
240 ((x >> 24) & 0x0000000000ff0000ULL) |
241 ((x >> 40) & 0x000000000000ff00ULL) |
242 ((x >> 56) & 0x00000000000000ffULL);
243 }
244 #endif
245
246
247 /* *************************************
248 * Architecture Macros
249 ***************************************/
250 typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
251
252 /* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */
253 #ifndef XXH_CPU_LITTLE_ENDIAN
254 static const int g_one = 1;
255 # define XXH_CPU_LITTLE_ENDIAN (*(const char*)(&g_one))
256 #endif
257
258
259 /* ***************************
260 * Memory reads
261 *****************************/
262 typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment;
263
XXH_readLE32_align(const void * ptr,XXH_endianess endian,XXH_alignment align)264 FORCE_INLINE_TEMPLATE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
265 {
266 if (align==XXH_unaligned)
267 return endian==XXH_littleEndian ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr));
268 else
269 return endian==XXH_littleEndian ? *(const U32*)ptr : XXH_swap32(*(const U32*)ptr);
270 }
271
XXH_readLE32(const void * ptr,XXH_endianess endian)272 FORCE_INLINE_TEMPLATE U32 XXH_readLE32(const void* ptr, XXH_endianess endian)
273 {
274 return XXH_readLE32_align(ptr, endian, XXH_unaligned);
275 }
276
XXH_readBE32(const void * ptr)277 static U32 XXH_readBE32(const void* ptr)
278 {
279 return XXH_CPU_LITTLE_ENDIAN ? XXH_swap32(XXH_read32(ptr)) : XXH_read32(ptr);
280 }
281
XXH_readLE64_align(const void * ptr,XXH_endianess endian,XXH_alignment align)282 FORCE_INLINE_TEMPLATE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
283 {
284 if (align==XXH_unaligned)
285 return endian==XXH_littleEndian ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr));
286 else
287 return endian==XXH_littleEndian ? *(const U64*)ptr : XXH_swap64(*(const U64*)ptr);
288 }
289
XXH_readLE64(const void * ptr,XXH_endianess endian)290 FORCE_INLINE_TEMPLATE U64 XXH_readLE64(const void* ptr, XXH_endianess endian)
291 {
292 return XXH_readLE64_align(ptr, endian, XXH_unaligned);
293 }
294
XXH_readBE64(const void * ptr)295 static U64 XXH_readBE64(const void* ptr)
296 {
297 return XXH_CPU_LITTLE_ENDIAN ? XXH_swap64(XXH_read64(ptr)) : XXH_read64(ptr);
298 }
299
300
301 /* *************************************
302 * Macros
303 ***************************************/
304 #define XXH_STATIC_ASSERT(c) { enum { XXH_static_assert = 1/(int)(!!(c)) }; } /* use only *after* variable declarations */
305
306
307 /* *************************************
308 * Constants
309 ***************************************/
310 static const U32 PRIME32_1 = 2654435761U;
311 static const U32 PRIME32_2 = 2246822519U;
312 static const U32 PRIME32_3 = 3266489917U;
313 static const U32 PRIME32_4 = 668265263U;
314 static const U32 PRIME32_5 = 374761393U;
315
316 static const U64 PRIME64_1 = 11400714785074694791ULL;
317 static const U64 PRIME64_2 = 14029467366897019727ULL;
318 static const U64 PRIME64_3 = 1609587929392839161ULL;
319 static const U64 PRIME64_4 = 9650029242287828579ULL;
320 static const U64 PRIME64_5 = 2870177450012600261ULL;
321
XXH_versionNumber(void)322 XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; }
323
324
325 /* **************************
326 * Utils
327 ****************************/
XXH32_copyState(XXH32_state_t * restrict dstState,const XXH32_state_t * restrict srcState)328 XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* restrict dstState, const XXH32_state_t* restrict srcState)
329 {
330 memcpy(dstState, srcState, sizeof(*dstState));
331 }
332
XXH64_copyState(XXH64_state_t * restrict dstState,const XXH64_state_t * restrict srcState)333 XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* restrict dstState, const XXH64_state_t* restrict srcState)
334 {
335 memcpy(dstState, srcState, sizeof(*dstState));
336 }
337
338
339 /* ***************************
340 * Simple Hash Functions
341 *****************************/
342
XXH32_round(U32 seed,U32 input)343 static U32 XXH32_round(U32 seed, U32 input)
344 {
345 seed += input * PRIME32_2;
346 seed = XXH_rotl32(seed, 13);
347 seed *= PRIME32_1;
348 return seed;
349 }
350
XXH32_endian_align(const void * input,size_t len,U32 seed,XXH_endianess endian,XXH_alignment align)351 FORCE_INLINE_TEMPLATE U32 XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_endianess endian, XXH_alignment align)
352 {
353 const BYTE* p = (const BYTE*)input;
354 const BYTE* bEnd = p + len;
355 U32 h32;
356 #define XXH_get32bits(p) XXH_readLE32_align(p, endian, align)
357
358 #ifdef XXH_ACCEPT_NULL_INPUT_POINTER
359 if (p==NULL) {
360 len=0;
361 bEnd=p=(const BYTE*)(size_t)16;
362 }
363 #endif
364
365 if (len>=16) {
366 const BYTE* const limit = bEnd - 16;
367 U32 v1 = seed + PRIME32_1 + PRIME32_2;
368 U32 v2 = seed + PRIME32_2;
369 U32 v3 = seed + 0;
370 U32 v4 = seed - PRIME32_1;
371
372 do {
373 v1 = XXH32_round(v1, XXH_get32bits(p)); p+=4;
374 v2 = XXH32_round(v2, XXH_get32bits(p)); p+=4;
375 v3 = XXH32_round(v3, XXH_get32bits(p)); p+=4;
376 v4 = XXH32_round(v4, XXH_get32bits(p)); p+=4;
377 } while (p<=limit);
378
379 h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
380 } else {
381 h32 = seed + PRIME32_5;
382 }
383
384 h32 += (U32) len;
385
386 while (p+4<=bEnd) {
387 h32 += XXH_get32bits(p) * PRIME32_3;
388 h32 = XXH_rotl32(h32, 17) * PRIME32_4 ;
389 p+=4;
390 }
391
392 while (p<bEnd) {
393 h32 += (*p) * PRIME32_5;
394 h32 = XXH_rotl32(h32, 11) * PRIME32_1 ;
395 p++;
396 }
397
398 h32 ^= h32 >> 15;
399 h32 *= PRIME32_2;
400 h32 ^= h32 >> 13;
401 h32 *= PRIME32_3;
402 h32 ^= h32 >> 16;
403
404 return h32;
405 }
406
407
XXH32(const void * input,size_t len,unsigned int seed)408 XXH_PUBLIC_API unsigned int XXH32 (const void* input, size_t len, unsigned int seed)
409 {
410 #if 0
411 /* Simple version, good for code maintenance, but unfortunately slow for small inputs */
412 XXH32_CREATESTATE_STATIC(state);
413 XXH32_reset(state, seed);
414 XXH32_update(state, input, len);
415 return XXH32_digest(state);
416 #else
417 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
418
419 if (XXH_FORCE_ALIGN_CHECK) {
420 if ((((size_t)input) & 3) == 0) { /* Input is 4-bytes aligned, leverage the speed benefit */
421 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
422 return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
423 else
424 return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
425 } }
426
427 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
428 return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
429 else
430 return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
431 #endif
432 }
433
434
XXH64_round(U64 acc,U64 input)435 static U64 XXH64_round(U64 acc, U64 input)
436 {
437 acc += input * PRIME64_2;
438 acc = XXH_rotl64(acc, 31);
439 acc *= PRIME64_1;
440 return acc;
441 }
442
XXH64_mergeRound(U64 acc,U64 val)443 static U64 XXH64_mergeRound(U64 acc, U64 val)
444 {
445 val = XXH64_round(0, val);
446 acc ^= val;
447 acc = acc * PRIME64_1 + PRIME64_4;
448 return acc;
449 }
450
XXH64_endian_align(const void * input,size_t len,U64 seed,XXH_endianess endian,XXH_alignment align)451 FORCE_INLINE_TEMPLATE U64 XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_endianess endian, XXH_alignment align)
452 {
453 const BYTE* p = (const BYTE*)input;
454 const BYTE* const bEnd = p + len;
455 U64 h64;
456 #define XXH_get64bits(p) XXH_readLE64_align(p, endian, align)
457
458 #ifdef XXH_ACCEPT_NULL_INPUT_POINTER
459 if (p==NULL) {
460 len=0;
461 bEnd=p=(const BYTE*)(size_t)32;
462 }
463 #endif
464
465 if (len>=32) {
466 const BYTE* const limit = bEnd - 32;
467 U64 v1 = seed + PRIME64_1 + PRIME64_2;
468 U64 v2 = seed + PRIME64_2;
469 U64 v3 = seed + 0;
470 U64 v4 = seed - PRIME64_1;
471
472 do {
473 v1 = XXH64_round(v1, XXH_get64bits(p)); p+=8;
474 v2 = XXH64_round(v2, XXH_get64bits(p)); p+=8;
475 v3 = XXH64_round(v3, XXH_get64bits(p)); p+=8;
476 v4 = XXH64_round(v4, XXH_get64bits(p)); p+=8;
477 } while (p<=limit);
478
479 h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
480 h64 = XXH64_mergeRound(h64, v1);
481 h64 = XXH64_mergeRound(h64, v2);
482 h64 = XXH64_mergeRound(h64, v3);
483 h64 = XXH64_mergeRound(h64, v4);
484
485 } else {
486 h64 = seed + PRIME64_5;
487 }
488
489 h64 += (U64) len;
490
491 while (p+8<=bEnd) {
492 U64 const k1 = XXH64_round(0, XXH_get64bits(p));
493 h64 ^= k1;
494 h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
495 p+=8;
496 }
497
498 if (p+4<=bEnd) {
499 h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1;
500 h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
501 p+=4;
502 }
503
504 while (p<bEnd) {
505 h64 ^= (*p) * PRIME64_5;
506 h64 = XXH_rotl64(h64, 11) * PRIME64_1;
507 p++;
508 }
509
510 h64 ^= h64 >> 33;
511 h64 *= PRIME64_2;
512 h64 ^= h64 >> 29;
513 h64 *= PRIME64_3;
514 h64 ^= h64 >> 32;
515
516 return h64;
517 }
518
519
XXH64(const void * input,size_t len,unsigned long long seed)520 XXH_PUBLIC_API unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed)
521 {
522 #if 0
523 /* Simple version, good for code maintenance, but unfortunately slow for small inputs */
524 XXH64_CREATESTATE_STATIC(state);
525 XXH64_reset(state, seed);
526 XXH64_update(state, input, len);
527 return XXH64_digest(state);
528 #else
529 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
530
531 if (XXH_FORCE_ALIGN_CHECK) {
532 if ((((size_t)input) & 7)==0) { /* Input is aligned, let's leverage the speed advantage */
533 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
534 return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
535 else
536 return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
537 } }
538
539 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
540 return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
541 else
542 return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
543 #endif
544 }
545
546
547 /* **************************************************
548 * Advanced Hash Functions
549 ****************************************************/
550
XXH32_createState(void)551 XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void)
552 {
553 return (XXH32_state_t*)XXH_malloc(sizeof(XXH32_state_t));
554 }
XXH32_freeState(XXH32_state_t * statePtr)555 XXH_PUBLIC_API XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr)
556 {
557 XXH_free(statePtr);
558 return XXH_OK;
559 }
560
XXH64_createState(void)561 XXH_PUBLIC_API XXH64_state_t* XXH64_createState(void)
562 {
563 return (XXH64_state_t*)XXH_malloc(sizeof(XXH64_state_t));
564 }
XXH64_freeState(XXH64_state_t * statePtr)565 XXH_PUBLIC_API XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr)
566 {
567 XXH_free(statePtr);
568 return XXH_OK;
569 }
570
571
572 /*** Hash feed ***/
573
XXH32_reset(XXH32_state_t * statePtr,unsigned int seed)574 XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, unsigned int seed)
575 {
576 XXH32_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
577 memset(&state, 0, sizeof(state)-4); /* do not write into reserved, for future removal */
578 state.v1 = seed + PRIME32_1 + PRIME32_2;
579 state.v2 = seed + PRIME32_2;
580 state.v3 = seed + 0;
581 state.v4 = seed - PRIME32_1;
582 memcpy(statePtr, &state, sizeof(state));
583 return XXH_OK;
584 }
585
586
XXH64_reset(XXH64_state_t * statePtr,unsigned long long seed)587 XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long long seed)
588 {
589 XXH64_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
590 memset(&state, 0, sizeof(state)-8); /* do not write into reserved, for future removal */
591 state.v1 = seed + PRIME64_1 + PRIME64_2;
592 state.v2 = seed + PRIME64_2;
593 state.v3 = seed + 0;
594 state.v4 = seed - PRIME64_1;
595 memcpy(statePtr, &state, sizeof(state));
596 return XXH_OK;
597 }
598
599
XXH32_update_endian(XXH32_state_t * state,const void * input,size_t len,XXH_endianess endian)600 FORCE_INLINE_TEMPLATE XXH_errorcode XXH32_update_endian (XXH32_state_t* state, const void* input, size_t len, XXH_endianess endian)
601 {
602 const BYTE* p = (const BYTE*)input;
603 const BYTE* const bEnd = p + len;
604
605 #ifdef XXH_ACCEPT_NULL_INPUT_POINTER
606 if (input==NULL) return XXH_ERROR;
607 #endif
608
609 state->total_len_32 += (unsigned)len;
610 state->large_len |= (len>=16) | (state->total_len_32>=16);
611
612 if (state->memsize + len < 16) { /* fill in tmp buffer */
613 XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len);
614 state->memsize += (unsigned)len;
615 return XXH_OK;
616 }
617
618 if (state->memsize) { /* some data left from previous update */
619 XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize);
620 { const U32* p32 = state->mem32;
621 state->v1 = XXH32_round(state->v1, XXH_readLE32(p32, endian)); p32++;
622 state->v2 = XXH32_round(state->v2, XXH_readLE32(p32, endian)); p32++;
623 state->v3 = XXH32_round(state->v3, XXH_readLE32(p32, endian)); p32++;
624 state->v4 = XXH32_round(state->v4, XXH_readLE32(p32, endian)); p32++;
625 }
626 p += 16-state->memsize;
627 state->memsize = 0;
628 }
629
630 if (p <= bEnd-16) {
631 const BYTE* const limit = bEnd - 16;
632 U32 v1 = state->v1;
633 U32 v2 = state->v2;
634 U32 v3 = state->v3;
635 U32 v4 = state->v4;
636
637 do {
638 v1 = XXH32_round(v1, XXH_readLE32(p, endian)); p+=4;
639 v2 = XXH32_round(v2, XXH_readLE32(p, endian)); p+=4;
640 v3 = XXH32_round(v3, XXH_readLE32(p, endian)); p+=4;
641 v4 = XXH32_round(v4, XXH_readLE32(p, endian)); p+=4;
642 } while (p<=limit);
643
644 state->v1 = v1;
645 state->v2 = v2;
646 state->v3 = v3;
647 state->v4 = v4;
648 }
649
650 if (p < bEnd) {
651 XXH_memcpy(state->mem32, p, (size_t)(bEnd-p));
652 state->memsize = (unsigned)(bEnd-p);
653 }
654
655 return XXH_OK;
656 }
657
XXH32_update(XXH32_state_t * state_in,const void * input,size_t len)658 XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len)
659 {
660 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
661
662 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
663 return XXH32_update_endian(state_in, input, len, XXH_littleEndian);
664 else
665 return XXH32_update_endian(state_in, input, len, XXH_bigEndian);
666 }
667
668
669
XXH32_digest_endian(const XXH32_state_t * state,XXH_endianess endian)670 FORCE_INLINE_TEMPLATE U32 XXH32_digest_endian (const XXH32_state_t* state, XXH_endianess endian)
671 {
672 const BYTE * p = (const BYTE*)state->mem32;
673 const BYTE* const bEnd = (const BYTE*)(state->mem32) + state->memsize;
674 U32 h32;
675
676 if (state->large_len) {
677 h32 = XXH_rotl32(state->v1, 1) + XXH_rotl32(state->v2, 7) + XXH_rotl32(state->v3, 12) + XXH_rotl32(state->v4, 18);
678 } else {
679 h32 = state->v3 /* == seed */ + PRIME32_5;
680 }
681
682 h32 += state->total_len_32;
683
684 while (p+4<=bEnd) {
685 h32 += XXH_readLE32(p, endian) * PRIME32_3;
686 h32 = XXH_rotl32(h32, 17) * PRIME32_4;
687 p+=4;
688 }
689
690 while (p<bEnd) {
691 h32 += (*p) * PRIME32_5;
692 h32 = XXH_rotl32(h32, 11) * PRIME32_1;
693 p++;
694 }
695
696 h32 ^= h32 >> 15;
697 h32 *= PRIME32_2;
698 h32 ^= h32 >> 13;
699 h32 *= PRIME32_3;
700 h32 ^= h32 >> 16;
701
702 return h32;
703 }
704
705
XXH32_digest(const XXH32_state_t * state_in)706 XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state_in)
707 {
708 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
709
710 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
711 return XXH32_digest_endian(state_in, XXH_littleEndian);
712 else
713 return XXH32_digest_endian(state_in, XXH_bigEndian);
714 }
715
716
717
718 /* **** XXH64 **** */
719
XXH64_update_endian(XXH64_state_t * state,const void * input,size_t len,XXH_endianess endian)720 FORCE_INLINE_TEMPLATE XXH_errorcode XXH64_update_endian (XXH64_state_t* state, const void* input, size_t len, XXH_endianess endian)
721 {
722 const BYTE* p = (const BYTE*)input;
723 const BYTE* const bEnd = p + len;
724
725 #ifdef XXH_ACCEPT_NULL_INPUT_POINTER
726 if (input==NULL) return XXH_ERROR;
727 #endif
728
729 state->total_len += len;
730
731 if (state->memsize + len < 32) { /* fill in tmp buffer */
732 XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
733 state->memsize += (U32)len;
734 return XXH_OK;
735 }
736
737 if (state->memsize) { /* tmp buffer is full */
738 XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize);
739 state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0, endian));
740 state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1, endian));
741 state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2, endian));
742 state->v4 = XXH64_round(state->v4, XXH_readLE64(state->mem64+3, endian));
743 p += 32-state->memsize;
744 state->memsize = 0;
745 }
746
747 if (p+32 <= bEnd) {
748 const BYTE* const limit = bEnd - 32;
749 U64 v1 = state->v1;
750 U64 v2 = state->v2;
751 U64 v3 = state->v3;
752 U64 v4 = state->v4;
753
754 do {
755 v1 = XXH64_round(v1, XXH_readLE64(p, endian)); p+=8;
756 v2 = XXH64_round(v2, XXH_readLE64(p, endian)); p+=8;
757 v3 = XXH64_round(v3, XXH_readLE64(p, endian)); p+=8;
758 v4 = XXH64_round(v4, XXH_readLE64(p, endian)); p+=8;
759 } while (p<=limit);
760
761 state->v1 = v1;
762 state->v2 = v2;
763 state->v3 = v3;
764 state->v4 = v4;
765 }
766
767 if (p < bEnd) {
768 XXH_memcpy(state->mem64, p, (size_t)(bEnd-p));
769 state->memsize = (unsigned)(bEnd-p);
770 }
771
772 return XXH_OK;
773 }
774
XXH64_update(XXH64_state_t * state_in,const void * input,size_t len)775 XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void* input, size_t len)
776 {
777 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
778
779 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
780 return XXH64_update_endian(state_in, input, len, XXH_littleEndian);
781 else
782 return XXH64_update_endian(state_in, input, len, XXH_bigEndian);
783 }
784
785
786
XXH64_digest_endian(const XXH64_state_t * state,XXH_endianess endian)787 FORCE_INLINE_TEMPLATE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endianess endian)
788 {
789 const BYTE * p = (const BYTE*)state->mem64;
790 const BYTE* const bEnd = (const BYTE*)state->mem64 + state->memsize;
791 U64 h64;
792
793 if (state->total_len >= 32) {
794 U64 const v1 = state->v1;
795 U64 const v2 = state->v2;
796 U64 const v3 = state->v3;
797 U64 const v4 = state->v4;
798
799 h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
800 h64 = XXH64_mergeRound(h64, v1);
801 h64 = XXH64_mergeRound(h64, v2);
802 h64 = XXH64_mergeRound(h64, v3);
803 h64 = XXH64_mergeRound(h64, v4);
804 } else {
805 h64 = state->v3 + PRIME64_5;
806 }
807
808 h64 += (U64) state->total_len;
809
810 while (p+8<=bEnd) {
811 U64 const k1 = XXH64_round(0, XXH_readLE64(p, endian));
812 h64 ^= k1;
813 h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
814 p+=8;
815 }
816
817 if (p+4<=bEnd) {
818 h64 ^= (U64)(XXH_readLE32(p, endian)) * PRIME64_1;
819 h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
820 p+=4;
821 }
822
823 while (p<bEnd) {
824 h64 ^= (*p) * PRIME64_5;
825 h64 = XXH_rotl64(h64, 11) * PRIME64_1;
826 p++;
827 }
828
829 h64 ^= h64 >> 33;
830 h64 *= PRIME64_2;
831 h64 ^= h64 >> 29;
832 h64 *= PRIME64_3;
833 h64 ^= h64 >> 32;
834
835 return h64;
836 }
837
838
XXH64_digest(const XXH64_state_t * state_in)839 XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state_in)
840 {
841 XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
842
843 if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
844 return XXH64_digest_endian(state_in, XXH_littleEndian);
845 else
846 return XXH64_digest_endian(state_in, XXH_bigEndian);
847 }
848
849
850 /* **************************
851 * Canonical representation
852 ****************************/
853
854 /*! Default XXH result types are basic unsigned 32 and 64 bits.
855 * The canonical representation follows human-readable write convention, aka big-endian (large digits first).
856 * These functions allow transformation of hash result into and from its canonical format.
857 * This way, hash values can be written into a file or buffer, and remain comparable across different systems and programs.
858 */
859
XXH32_canonicalFromHash(XXH32_canonical_t * dst,XXH32_hash_t hash)860 XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash)
861 {
862 XXH_STATIC_ASSERT(sizeof(XXH32_canonical_t) == sizeof(XXH32_hash_t));
863 if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap32(hash);
864 memcpy(dst, &hash, sizeof(*dst));
865 }
866
XXH64_canonicalFromHash(XXH64_canonical_t * dst,XXH64_hash_t hash)867 XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t hash)
868 {
869 XXH_STATIC_ASSERT(sizeof(XXH64_canonical_t) == sizeof(XXH64_hash_t));
870 if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap64(hash);
871 memcpy(dst, &hash, sizeof(*dst));
872 }
873
XXH32_hashFromCanonical(const XXH32_canonical_t * src)874 XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src)
875 {
876 return XXH_readBE32(src);
877 }
878
XXH64_hashFromCanonical(const XXH64_canonical_t * src)879 XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src)
880 {
881 return XXH_readBE64(src);
882 }
883