• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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__) \
54                         || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) \
55                         || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
56 #    define XXH_FORCE_MEMORY_ACCESS 2
57 #  elif (defined(__INTEL_COMPILER) && !defined(_WIN32)) || \
58   (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) \
59                     || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) \
60                     || defined(__ARM_ARCH_7S__) ))
61 #    define XXH_FORCE_MEMORY_ACCESS 1
62 #  endif
63 #endif
64 
65 /*!XXH_ACCEPT_NULL_INPUT_POINTER :
66  * If input pointer is NULL, xxHash default behavior is to dereference it, triggering a segfault.
67  * When this macro is enabled, xxHash actively checks input for null pointer.
68  * It it is, result for null input pointers is the same as a null-length input.
69  */
70 #ifndef XXH_ACCEPT_NULL_INPUT_POINTER   /* can be defined externally */
71 #  define XXH_ACCEPT_NULL_INPUT_POINTER 0
72 #endif
73 
74 /*!XXH_FORCE_NATIVE_FORMAT :
75  * By default, xxHash library provides endian-independent Hash values, based on little-endian convention.
76  * Results are therefore identical for little-endian and big-endian CPU.
77  * This comes at a performance cost for big-endian CPU, since some swapping is required to emulate little-endian format.
78  * Should endian-independence be of no importance for your application, you may set the #define below to 1,
79  * to improve speed for Big-endian CPU.
80  * This option has no impact on Little_Endian CPU.
81  */
82 #ifndef XXH_FORCE_NATIVE_FORMAT   /* can be defined externally */
83 #  define XXH_FORCE_NATIVE_FORMAT 0
84 #endif
85 
86 /*!XXH_FORCE_ALIGN_CHECK :
87  * This is a minor performance trick, only useful with lots of very small keys.
88  * It means : check for aligned/unaligned input.
89  * The check costs one initial branch per hash;
90  * set it to 0 when the input is guaranteed to be aligned,
91  * or when alignment doesn't matter for performance.
92  */
93 #ifndef XXH_FORCE_ALIGN_CHECK /* can be defined externally */
94 #  if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64)
95 #    define XXH_FORCE_ALIGN_CHECK 0
96 #  else
97 #    define XXH_FORCE_ALIGN_CHECK 1
98 #  endif
99 #endif
100 
101 
102 /* *************************************
103 *  Includes & Memory related functions
104 ***************************************/
105 /*! Modify the local functions below should you wish to use some other memory routines
106 *   for malloc(), free() */
107 #include <stdlib.h>
XXH_malloc(size_t s)108 static void* XXH_malloc(size_t s) { return malloc(s); }
XXH_free(void * p)109 static void  XXH_free  (void* p)  { free(p); }
110 /*! and for memcpy() */
111 #include <string.h>
XXH_memcpy(void * dest,const void * src,size_t size)112 static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
113 
114 #include <assert.h>   /* assert */
115 
116 #define XXH_STATIC_LINKING_ONLY
117 #include "xxhash.h"
118 
119 
120 /* *************************************
121 *  Compiler Specific Options
122 ***************************************/
123 #ifdef _MSC_VER    /* Visual Studio */
124 #  pragma warning(disable : 4127)      /* disable: C4127: conditional expression is constant */
125 #  define FORCE_INLINE static __forceinline
126 #else
127 #  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
128 #    ifdef __GNUC__
129 #      define FORCE_INLINE static inline __attribute__((always_inline))
130 #    else
131 #      define FORCE_INLINE static inline
132 #    endif
133 #  else
134 #    define FORCE_INLINE static
135 #  endif /* __STDC_VERSION__ */
136 #endif
137 
138 
139 /* *************************************
140 *  Basic Types
141 ***************************************/
142 #ifndef MEM_MODULE
143 # if !defined (__VMS) \
144   && (defined (__cplusplus) \
145   || (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 # else
151     typedef unsigned char      BYTE;
152     typedef unsigned short     U16;
153     typedef unsigned int       U32;
154 # endif
155 #endif
156 
157 #if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2))
158 
159 /* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */
XXH_read32(const void * memPtr)160 static U32 XXH_read32(const void* memPtr) { return *(const U32*) memPtr; }
161 
162 #elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1))
163 
164 /* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
165 /* currently only defined for gcc and icc */
166 typedef union { U32 u32; } __attribute__((packed)) unalign;
XXH_read32(const void * ptr)167 static U32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
168 
169 #else
170 
171 /* portable and safe solution. Generally efficient.
172  * see : http://stackoverflow.com/a/32095106/646947
173  */
XXH_read32(const void * memPtr)174 static U32 XXH_read32(const void* memPtr)
175 {
176     U32 val;
177     memcpy(&val, memPtr, sizeof(val));
178     return val;
179 }
180 
181 #endif   /* XXH_FORCE_DIRECT_MEMORY_ACCESS */
182 
183 
184 /* ****************************************
185 *  Compiler-specific Functions and Macros
186 ******************************************/
187 #define XXH_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
188 
189 /* Note : although _rotl exists for minGW (GCC under windows), performance seems poor */
190 #if defined(_MSC_VER)
191 #  define XXH_rotl32(x,r) _rotl(x,r)
192 #  define XXH_rotl64(x,r) _rotl64(x,r)
193 #else
194 #  define XXH_rotl32(x,r) ((x << r) | (x >> (32 - r)))
195 #  define XXH_rotl64(x,r) ((x << r) | (x >> (64 - r)))
196 #endif
197 
198 #if defined(_MSC_VER)     /* Visual Studio */
199 #  define XXH_swap32 _byteswap_ulong
200 #elif XXH_GCC_VERSION >= 403
201 #  define XXH_swap32 __builtin_bswap32
202 #else
XXH_swap32(U32 x)203 static U32 XXH_swap32 (U32 x)
204 {
205     return  ((x << 24) & 0xff000000 ) |
206             ((x <<  8) & 0x00ff0000 ) |
207             ((x >>  8) & 0x0000ff00 ) |
208             ((x >> 24) & 0x000000ff );
209 }
210 #endif
211 
212 
213 /* *************************************
214 *  Architecture Macros
215 ***************************************/
216 typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
217 
218 /* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */
219 #ifndef XXH_CPU_LITTLE_ENDIAN
XXH_isLittleEndian(void)220 static int XXH_isLittleEndian(void)
221 {
222     const union { U32 u; BYTE c[4]; } one = { 1 };   /* don't use static : performance detrimental  */
223     return one.c[0];
224 }
225 #   define XXH_CPU_LITTLE_ENDIAN   XXH_isLittleEndian()
226 #endif
227 
228 
229 /* ***************************
230 *  Memory reads
231 *****************************/
232 typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment;
233 
XXH_readLE32_align(const void * ptr,XXH_endianess endian,XXH_alignment align)234 FORCE_INLINE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
235 {
236     if (align==XXH_unaligned)
237         return endian==XXH_littleEndian ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr));
238     else
239         return endian==XXH_littleEndian ? *(const U32*)ptr : XXH_swap32(*(const U32*)ptr);
240 }
241 
XXH_readLE32(const void * ptr,XXH_endianess endian)242 FORCE_INLINE U32 XXH_readLE32(const void* ptr, XXH_endianess endian)
243 {
244     return XXH_readLE32_align(ptr, endian, XXH_unaligned);
245 }
246 
XXH_readBE32(const void * ptr)247 static U32 XXH_readBE32(const void* ptr)
248 {
249     return XXH_CPU_LITTLE_ENDIAN ? XXH_swap32(XXH_read32(ptr)) : XXH_read32(ptr);
250 }
251 
252 
253 /* *************************************
254 *  Macros
255 ***************************************/
256 #define XXH_STATIC_ASSERT(c)  { enum { XXH_sa = 1/(int)(!!(c)) }; }  /* use after variable declarations */
XXH_versionNumber(void)257 XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; }
258 
259 
260 /* *******************************************************************
261 *  32-bit hash functions
262 *********************************************************************/
263 static const U32 PRIME32_1 = 2654435761U;
264 static const U32 PRIME32_2 = 2246822519U;
265 static const U32 PRIME32_3 = 3266489917U;
266 static const U32 PRIME32_4 =  668265263U;
267 static const U32 PRIME32_5 =  374761393U;
268 
XXH32_round(U32 seed,U32 input)269 static U32 XXH32_round(U32 seed, U32 input)
270 {
271     seed += input * PRIME32_2;
272     seed  = XXH_rotl32(seed, 13);
273     seed *= PRIME32_1;
274     return seed;
275 }
276 
277 /* mix all bits */
XXH32_avalanche(U32 h32)278 static U32 XXH32_avalanche(U32 h32)
279 {
280     h32 ^= h32 >> 15;
281     h32 *= PRIME32_2;
282     h32 ^= h32 >> 13;
283     h32 *= PRIME32_3;
284     h32 ^= h32 >> 16;
285     return(h32);
286 }
287 
288 #define XXH_get32bits(p) XXH_readLE32_align(p, endian, align)
289 
290 static U32
XXH32_finalize(U32 h32,const void * ptr,size_t len,XXH_endianess endian,XXH_alignment align)291 XXH32_finalize(U32 h32, const void* ptr, size_t len,
292                 XXH_endianess endian, XXH_alignment align)
293 
294 {
295     const BYTE* p = (const BYTE*)ptr;
296 
297 #define PROCESS1               \
298     h32 += (*p++) * PRIME32_5; \
299     h32 = XXH_rotl32(h32, 11) * PRIME32_1 ;
300 
301 #define PROCESS4                         \
302     h32 += XXH_get32bits(p) * PRIME32_3; \
303     p+=4;                                \
304     h32  = XXH_rotl32(h32, 17) * PRIME32_4 ;
305 
306     switch(len&15)  /* or switch(bEnd - p) */
307     {
308       case 12:      PROCESS4;
309                     /* fallthrough */
310       case 8:       PROCESS4;
311                     /* fallthrough */
312       case 4:       PROCESS4;
313                     return XXH32_avalanche(h32);
314 
315       case 13:      PROCESS4;
316                     /* fallthrough */
317       case 9:       PROCESS4;
318                     /* fallthrough */
319       case 5:       PROCESS4;
320                     PROCESS1;
321                     return XXH32_avalanche(h32);
322 
323       case 14:      PROCESS4;
324                     /* fallthrough */
325       case 10:      PROCESS4;
326                     /* fallthrough */
327       case 6:       PROCESS4;
328                     PROCESS1;
329                     PROCESS1;
330                     return XXH32_avalanche(h32);
331 
332       case 15:      PROCESS4;
333                     /* fallthrough */
334       case 11:      PROCESS4;
335                     /* fallthrough */
336       case 7:       PROCESS4;
337                     /* fallthrough */
338       case 3:       PROCESS1;
339                     /* fallthrough */
340       case 2:       PROCESS1;
341                     /* fallthrough */
342       case 1:       PROCESS1;
343                     /* fallthrough */
344       case 0:       return XXH32_avalanche(h32);
345     }
346     assert(0);
347     return h32;   /* reaching this point is deemed impossible */
348 }
349 
350 
351 FORCE_INLINE U32
XXH32_endian_align(const void * input,size_t len,U32 seed,XXH_endianess endian,XXH_alignment align)352 XXH32_endian_align(const void* input, size_t len, U32 seed,
353                     XXH_endianess endian, XXH_alignment align)
354 {
355     const BYTE* p = (const BYTE*)input;
356     const BYTE* bEnd = p + len;
357     U32 h32;
358 
359 #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
360     if (p==NULL) {
361         len=0;
362         bEnd=p=(const BYTE*)(size_t)16;
363     }
364 #endif
365 
366     if (len>=16) {
367         const BYTE* const limit = bEnd - 15;
368         U32 v1 = seed + PRIME32_1 + PRIME32_2;
369         U32 v2 = seed + PRIME32_2;
370         U32 v3 = seed + 0;
371         U32 v4 = seed - PRIME32_1;
372 
373         do {
374             v1 = XXH32_round(v1, XXH_get32bits(p)); p+=4;
375             v2 = XXH32_round(v2, XXH_get32bits(p)); p+=4;
376             v3 = XXH32_round(v3, XXH_get32bits(p)); p+=4;
377             v4 = XXH32_round(v4, XXH_get32bits(p)); p+=4;
378         } while (p < limit);
379 
380         h32 = XXH_rotl32(v1, 1)  + XXH_rotl32(v2, 7)
381             + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
382     } else {
383         h32  = seed + PRIME32_5;
384     }
385 
386     h32 += (U32)len;
387 
388     return XXH32_finalize(h32, p, len&15, endian, align);
389 }
390 
391 
XXH32(const void * input,size_t len,unsigned int seed)392 XXH_PUBLIC_API unsigned int XXH32 (const void* input, size_t len, unsigned int seed)
393 {
394 #if 0
395     /* Simple version, good for code maintenance, but unfortunately slow for small inputs */
396     XXH32_state_t state;
397     XXH32_reset(&state, seed);
398     XXH32_update(&state, input, len);
399     return XXH32_digest(&state);
400 #else
401     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
402 
403     if (XXH_FORCE_ALIGN_CHECK) {
404         if ((((size_t)input) & 3) == 0) {   /* Input is 4-bytes aligned, leverage the speed benefit */
405             if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
406                 return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
407             else
408                 return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
409     }   }
410 
411     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
412         return XXH32_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
413     else
414         return XXH32_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
415 #endif
416 }
417 
418 
419 
420 /*======   Hash streaming   ======*/
421 
XXH32_createState(void)422 XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void)
423 {
424     return (XXH32_state_t*)XXH_malloc(sizeof(XXH32_state_t));
425 }
XXH32_freeState(XXH32_state_t * statePtr)426 XXH_PUBLIC_API XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr)
427 {
428     XXH_free(statePtr);
429     return XXH_OK;
430 }
431 
XXH32_copyState(XXH32_state_t * dstState,const XXH32_state_t * srcState)432 XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* dstState, const XXH32_state_t* srcState)
433 {
434     memcpy(dstState, srcState, sizeof(*dstState));
435 }
436 
XXH32_reset(XXH32_state_t * statePtr,unsigned int seed)437 XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, unsigned int seed)
438 {
439     XXH32_state_t state;   /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
440     memset(&state, 0, sizeof(state));
441     state.v1 = seed + PRIME32_1 + PRIME32_2;
442     state.v2 = seed + PRIME32_2;
443     state.v3 = seed + 0;
444     state.v4 = seed - PRIME32_1;
445     /* do not write into reserved, planned to be removed in a future version */
446     memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved));
447     return XXH_OK;
448 }
449 
450 
451 FORCE_INLINE XXH_errorcode
XXH32_update_endian(XXH32_state_t * state,const void * input,size_t len,XXH_endianess endian)452 XXH32_update_endian(XXH32_state_t* state, const void* input, size_t len, XXH_endianess endian)
453 {
454     if (input==NULL)
455 #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
456         return XXH_OK;
457 #else
458         return XXH_ERROR;
459 #endif
460 
461     {   const BYTE* p = (const BYTE*)input;
462         const BYTE* const bEnd = p + len;
463 
464         state->total_len_32 += (unsigned)len;
465         state->large_len |= (len>=16) | (state->total_len_32>=16);
466 
467         if (state->memsize + len < 16)  {   /* fill in tmp buffer */
468             XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len);
469             state->memsize += (unsigned)len;
470             return XXH_OK;
471         }
472 
473         if (state->memsize) {   /* some data left from previous update */
474             XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize);
475             {   const U32* p32 = state->mem32;
476                 state->v1 = XXH32_round(state->v1, XXH_readLE32(p32, endian)); p32++;
477                 state->v2 = XXH32_round(state->v2, XXH_readLE32(p32, endian)); p32++;
478                 state->v3 = XXH32_round(state->v3, XXH_readLE32(p32, endian)); p32++;
479                 state->v4 = XXH32_round(state->v4, XXH_readLE32(p32, endian));
480             }
481             p += 16-state->memsize;
482             state->memsize = 0;
483         }
484 
485         if (p <= bEnd-16) {
486             const BYTE* const limit = bEnd - 16;
487             U32 v1 = state->v1;
488             U32 v2 = state->v2;
489             U32 v3 = state->v3;
490             U32 v4 = state->v4;
491 
492             do {
493                 v1 = XXH32_round(v1, XXH_readLE32(p, endian)); p+=4;
494                 v2 = XXH32_round(v2, XXH_readLE32(p, endian)); p+=4;
495                 v3 = XXH32_round(v3, XXH_readLE32(p, endian)); p+=4;
496                 v4 = XXH32_round(v4, XXH_readLE32(p, endian)); p+=4;
497             } while (p<=limit);
498 
499             state->v1 = v1;
500             state->v2 = v2;
501             state->v3 = v3;
502             state->v4 = v4;
503         }
504 
505         if (p < bEnd) {
506             XXH_memcpy(state->mem32, p, (size_t)(bEnd-p));
507             state->memsize = (unsigned)(bEnd-p);
508         }
509     }
510 
511     return XXH_OK;
512 }
513 
514 
XXH32_update(XXH32_state_t * state_in,const void * input,size_t len)515 XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len)
516 {
517     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
518 
519     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
520         return XXH32_update_endian(state_in, input, len, XXH_littleEndian);
521     else
522         return XXH32_update_endian(state_in, input, len, XXH_bigEndian);
523 }
524 
525 
526 FORCE_INLINE U32
XXH32_digest_endian(const XXH32_state_t * state,XXH_endianess endian)527 XXH32_digest_endian (const XXH32_state_t* state, XXH_endianess endian)
528 {
529     U32 h32;
530 
531     if (state->large_len) {
532         h32 = XXH_rotl32(state->v1, 1)
533             + XXH_rotl32(state->v2, 7)
534             + XXH_rotl32(state->v3, 12)
535             + XXH_rotl32(state->v4, 18);
536     } else {
537         h32 = state->v3 /* == seed */ + PRIME32_5;
538     }
539 
540     h32 += state->total_len_32;
541 
542     return XXH32_finalize(h32, state->mem32, state->memsize, endian, XXH_aligned);
543 }
544 
545 
XXH32_digest(const XXH32_state_t * state_in)546 XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state_in)
547 {
548     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
549 
550     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
551         return XXH32_digest_endian(state_in, XXH_littleEndian);
552     else
553         return XXH32_digest_endian(state_in, XXH_bigEndian);
554 }
555 
556 
557 /*======   Canonical representation   ======*/
558 
559 /*! Default XXH result types are basic unsigned 32 and 64 bits.
560 *   The canonical representation follows human-readable write convention, aka big-endian (large digits first).
561 *   These functions allow transformation of hash result into and from its canonical format.
562 *   This way, hash values can be written into a file or buffer, remaining comparable across different systems.
563 */
564 
XXH32_canonicalFromHash(XXH32_canonical_t * dst,XXH32_hash_t hash)565 XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash)
566 {
567     XXH_STATIC_ASSERT(sizeof(XXH32_canonical_t) == sizeof(XXH32_hash_t));
568     if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap32(hash);
569     memcpy(dst, &hash, sizeof(*dst));
570 }
571 
XXH32_hashFromCanonical(const XXH32_canonical_t * src)572 XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src)
573 {
574     return XXH_readBE32(src);
575 }
576 
577 
578 #ifndef XXH_NO_LONG_LONG
579 
580 /* *******************************************************************
581 *  64-bit hash functions
582 *********************************************************************/
583 
584 /*======   Memory access   ======*/
585 
586 #ifndef MEM_MODULE
587 # define MEM_MODULE
588 # if !defined (__VMS) \
589   && (defined (__cplusplus) \
590   || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
591 #   include <stdint.h>
592     typedef uint64_t U64;
593 # else
594     /* if compiler doesn't support unsigned long long, replace by another 64-bit type */
595     typedef unsigned long long U64;
596 # endif
597 #endif
598 
599 
600 #if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2))
601 
602 /* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */
XXH_read64(const void * memPtr)603 static U64 XXH_read64(const void* memPtr) { return *(const U64*) memPtr; }
604 
605 #elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1))
606 
607 /* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
608 /* currently only defined for gcc and icc */
609 typedef union { U32 u32; U64 u64; } __attribute__((packed)) unalign64;
XXH_read64(const void * ptr)610 static U64 XXH_read64(const void* ptr) { return ((const unalign64*)ptr)->u64; }
611 
612 #else
613 
614 /* portable and safe solution. Generally efficient.
615  * see : http://stackoverflow.com/a/32095106/646947
616  */
617 
XXH_read64(const void * memPtr)618 static U64 XXH_read64(const void* memPtr)
619 {
620     U64 val;
621     memcpy(&val, memPtr, sizeof(val));
622     return val;
623 }
624 
625 #endif   /* XXH_FORCE_DIRECT_MEMORY_ACCESS */
626 
627 #if defined(_MSC_VER)     /* Visual Studio */
628 #  define XXH_swap64 _byteswap_uint64
629 #elif XXH_GCC_VERSION >= 403
630 #  define XXH_swap64 __builtin_bswap64
631 #else
XXH_swap64(U64 x)632 static U64 XXH_swap64 (U64 x)
633 {
634     return  ((x << 56) & 0xff00000000000000ULL) |
635             ((x << 40) & 0x00ff000000000000ULL) |
636             ((x << 24) & 0x0000ff0000000000ULL) |
637             ((x << 8)  & 0x000000ff00000000ULL) |
638             ((x >> 8)  & 0x00000000ff000000ULL) |
639             ((x >> 24) & 0x0000000000ff0000ULL) |
640             ((x >> 40) & 0x000000000000ff00ULL) |
641             ((x >> 56) & 0x00000000000000ffULL);
642 }
643 #endif
644 
XXH_readLE64_align(const void * ptr,XXH_endianess endian,XXH_alignment align)645 FORCE_INLINE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
646 {
647     if (align==XXH_unaligned)
648         return endian==XXH_littleEndian ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr));
649     else
650         return endian==XXH_littleEndian ? *(const U64*)ptr : XXH_swap64(*(const U64*)ptr);
651 }
652 
XXH_readLE64(const void * ptr,XXH_endianess endian)653 FORCE_INLINE U64 XXH_readLE64(const void* ptr, XXH_endianess endian)
654 {
655     return XXH_readLE64_align(ptr, endian, XXH_unaligned);
656 }
657 
XXH_readBE64(const void * ptr)658 static U64 XXH_readBE64(const void* ptr)
659 {
660     return XXH_CPU_LITTLE_ENDIAN ? XXH_swap64(XXH_read64(ptr)) : XXH_read64(ptr);
661 }
662 
663 
664 /*======   xxh64   ======*/
665 
666 static const U64 PRIME64_1 = 11400714785074694791ULL;
667 static const U64 PRIME64_2 = 14029467366897019727ULL;
668 static const U64 PRIME64_3 =  1609587929392839161ULL;
669 static const U64 PRIME64_4 =  9650029242287828579ULL;
670 static const U64 PRIME64_5 =  2870177450012600261ULL;
671 
XXH64_round(U64 acc,U64 input)672 static U64 XXH64_round(U64 acc, U64 input)
673 {
674     acc += input * PRIME64_2;
675     acc  = XXH_rotl64(acc, 31);
676     acc *= PRIME64_1;
677     return acc;
678 }
679 
XXH64_mergeRound(U64 acc,U64 val)680 static U64 XXH64_mergeRound(U64 acc, U64 val)
681 {
682     val  = XXH64_round(0, val);
683     acc ^= val;
684     acc  = acc * PRIME64_1 + PRIME64_4;
685     return acc;
686 }
687 
XXH64_avalanche(U64 h64)688 static U64 XXH64_avalanche(U64 h64)
689 {
690     h64 ^= h64 >> 33;
691     h64 *= PRIME64_2;
692     h64 ^= h64 >> 29;
693     h64 *= PRIME64_3;
694     h64 ^= h64 >> 32;
695     return h64;
696 }
697 
698 
699 #define XXH_get64bits(p) XXH_readLE64_align(p, endian, align)
700 
701 static U64
XXH64_finalize(U64 h64,const void * ptr,size_t len,XXH_endianess endian,XXH_alignment align)702 XXH64_finalize(U64 h64, const void* ptr, size_t len,
703                XXH_endianess endian, XXH_alignment align)
704 {
705     const BYTE* p = (const BYTE*)ptr;
706 
707 #define PROCESS1_64            \
708     h64 ^= (*p++) * PRIME64_5; \
709     h64 = XXH_rotl64(h64, 11) * PRIME64_1;
710 
711 #define PROCESS4_64          \
712     h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1; \
713     p+=4;                    \
714     h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
715 
716 #define PROCESS8_64 {        \
717     U64 const k1 = XXH64_round(0, XXH_get64bits(p)); \
718     p+=8;                    \
719     h64 ^= k1;               \
720     h64  = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; \
721 }
722 
723     switch(len&31) {
724       case 24: PROCESS8_64;
725                     /* fallthrough */
726       case 16: PROCESS8_64;
727                     /* fallthrough */
728       case  8: PROCESS8_64;
729                return XXH64_avalanche(h64);
730 
731       case 28: PROCESS8_64;
732                     /* fallthrough */
733       case 20: PROCESS8_64;
734                     /* fallthrough */
735       case 12: PROCESS8_64;
736                     /* fallthrough */
737       case  4: PROCESS4_64;
738                return XXH64_avalanche(h64);
739 
740       case 25: PROCESS8_64;
741                     /* fallthrough */
742       case 17: PROCESS8_64;
743                     /* fallthrough */
744       case  9: PROCESS8_64;
745                PROCESS1_64;
746                return XXH64_avalanche(h64);
747 
748       case 29: PROCESS8_64;
749                     /* fallthrough */
750       case 21: PROCESS8_64;
751                     /* fallthrough */
752       case 13: PROCESS8_64;
753                     /* fallthrough */
754       case  5: PROCESS4_64;
755                PROCESS1_64;
756                return XXH64_avalanche(h64);
757 
758       case 26: PROCESS8_64;
759                     /* fallthrough */
760       case 18: PROCESS8_64;
761                     /* fallthrough */
762       case 10: PROCESS8_64;
763                PROCESS1_64;
764                PROCESS1_64;
765                return XXH64_avalanche(h64);
766 
767       case 30: PROCESS8_64;
768                     /* fallthrough */
769       case 22: PROCESS8_64;
770                     /* fallthrough */
771       case 14: PROCESS8_64;
772                     /* fallthrough */
773       case  6: PROCESS4_64;
774                PROCESS1_64;
775                PROCESS1_64;
776                return XXH64_avalanche(h64);
777 
778       case 27: PROCESS8_64;
779                     /* fallthrough */
780       case 19: PROCESS8_64;
781                     /* fallthrough */
782       case 11: PROCESS8_64;
783                PROCESS1_64;
784                PROCESS1_64;
785                PROCESS1_64;
786                return XXH64_avalanche(h64);
787 
788       case 31: PROCESS8_64;
789                     /* fallthrough */
790       case 23: PROCESS8_64;
791                     /* fallthrough */
792       case 15: PROCESS8_64;
793                     /* fallthrough */
794       case  7: PROCESS4_64;
795                     /* fallthrough */
796       case  3: PROCESS1_64;
797                     /* fallthrough */
798       case  2: PROCESS1_64;
799                     /* fallthrough */
800       case  1: PROCESS1_64;
801                     /* fallthrough */
802       case  0: return XXH64_avalanche(h64);
803     }
804 
805     /* impossible to reach */
806     assert(0);
807     return 0;  /* unreachable, but some compilers complain without it */
808 }
809 
810 FORCE_INLINE U64
XXH64_endian_align(const void * input,size_t len,U64 seed,XXH_endianess endian,XXH_alignment align)811 XXH64_endian_align(const void* input, size_t len, U64 seed,
812                 XXH_endianess endian, XXH_alignment align)
813 {
814     const BYTE* p = (const BYTE*)input;
815     const BYTE* bEnd = p + len;
816     U64 h64;
817 
818 #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
819     if (p==NULL) {
820         len=0;
821         bEnd=p=(const BYTE*)(size_t)32;
822     }
823 #endif
824 
825     if (len>=32) {
826         const BYTE* const limit = bEnd - 32;
827         U64 v1 = seed + PRIME64_1 + PRIME64_2;
828         U64 v2 = seed + PRIME64_2;
829         U64 v3 = seed + 0;
830         U64 v4 = seed - PRIME64_1;
831 
832         do {
833             v1 = XXH64_round(v1, XXH_get64bits(p)); p+=8;
834             v2 = XXH64_round(v2, XXH_get64bits(p)); p+=8;
835             v3 = XXH64_round(v3, XXH_get64bits(p)); p+=8;
836             v4 = XXH64_round(v4, XXH_get64bits(p)); p+=8;
837         } while (p<=limit);
838 
839         h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
840         h64 = XXH64_mergeRound(h64, v1);
841         h64 = XXH64_mergeRound(h64, v2);
842         h64 = XXH64_mergeRound(h64, v3);
843         h64 = XXH64_mergeRound(h64, v4);
844 
845     } else {
846         h64  = seed + PRIME64_5;
847     }
848 
849     h64 += (U64) len;
850 
851     return XXH64_finalize(h64, p, len, endian, align);
852 }
853 
854 
XXH64(const void * input,size_t len,unsigned long long seed)855 XXH_PUBLIC_API unsigned long long XXH64 (const void* input, size_t len, unsigned long long seed)
856 {
857 #if 0
858     /* Simple version, good for code maintenance, but unfortunately slow for small inputs */
859     XXH64_state_t state;
860     XXH64_reset(&state, seed);
861     XXH64_update(&state, input, len);
862     return XXH64_digest(&state);
863 #else
864     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
865 
866     if (XXH_FORCE_ALIGN_CHECK) {
867         if ((((size_t)input) & 7)==0) {  /* Input is aligned, let's leverage the speed advantage */
868             if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
869                 return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_aligned);
870             else
871                 return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_aligned);
872     }   }
873 
874     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
875         return XXH64_endian_align(input, len, seed, XXH_littleEndian, XXH_unaligned);
876     else
877         return XXH64_endian_align(input, len, seed, XXH_bigEndian, XXH_unaligned);
878 #endif
879 }
880 
881 /*======   Hash Streaming   ======*/
882 
XXH64_createState(void)883 XXH_PUBLIC_API XXH64_state_t* XXH64_createState(void)
884 {
885     return (XXH64_state_t*)XXH_malloc(sizeof(XXH64_state_t));
886 }
XXH64_freeState(XXH64_state_t * statePtr)887 XXH_PUBLIC_API XXH_errorcode XXH64_freeState(XXH64_state_t* statePtr)
888 {
889     XXH_free(statePtr);
890     return XXH_OK;
891 }
892 
XXH64_copyState(XXH64_state_t * dstState,const XXH64_state_t * srcState)893 XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* dstState, const XXH64_state_t* srcState)
894 {
895     memcpy(dstState, srcState, sizeof(*dstState));
896 }
897 
XXH64_reset(XXH64_state_t * statePtr,unsigned long long seed)898 XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long long seed)
899 {
900     XXH64_state_t state;   /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
901     memset(&state, 0, sizeof(state));
902     state.v1 = seed + PRIME64_1 + PRIME64_2;
903     state.v2 = seed + PRIME64_2;
904     state.v3 = seed + 0;
905     state.v4 = seed - PRIME64_1;
906      /* do not write into reserved, planned to be removed in a future version */
907     memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved));
908     return XXH_OK;
909 }
910 
911 FORCE_INLINE XXH_errorcode
XXH64_update_endian(XXH64_state_t * state,const void * input,size_t len,XXH_endianess endian)912 XXH64_update_endian (XXH64_state_t* state, const void* input, size_t len, XXH_endianess endian)
913 {
914     if (input==NULL)
915 #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
916         return XXH_OK;
917 #else
918         return XXH_ERROR;
919 #endif
920 
921     {   const BYTE* p = (const BYTE*)input;
922         const BYTE* const bEnd = p + len;
923 
924         state->total_len += len;
925 
926         if (state->memsize + len < 32) {  /* fill in tmp buffer */
927             XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
928             state->memsize += (U32)len;
929             return XXH_OK;
930         }
931 
932         if (state->memsize) {   /* tmp buffer is full */
933             XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize);
934             state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0, endian));
935             state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1, endian));
936             state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2, endian));
937             state->v4 = XXH64_round(state->v4, XXH_readLE64(state->mem64+3, endian));
938             p += 32-state->memsize;
939             state->memsize = 0;
940         }
941 
942         if (p+32 <= bEnd) {
943             const BYTE* const limit = bEnd - 32;
944             U64 v1 = state->v1;
945             U64 v2 = state->v2;
946             U64 v3 = state->v3;
947             U64 v4 = state->v4;
948 
949             do {
950                 v1 = XXH64_round(v1, XXH_readLE64(p, endian)); p+=8;
951                 v2 = XXH64_round(v2, XXH_readLE64(p, endian)); p+=8;
952                 v3 = XXH64_round(v3, XXH_readLE64(p, endian)); p+=8;
953                 v4 = XXH64_round(v4, XXH_readLE64(p, endian)); p+=8;
954             } while (p<=limit);
955 
956             state->v1 = v1;
957             state->v2 = v2;
958             state->v3 = v3;
959             state->v4 = v4;
960         }
961 
962         if (p < bEnd) {
963             XXH_memcpy(state->mem64, p, (size_t)(bEnd-p));
964             state->memsize = (unsigned)(bEnd-p);
965         }
966     }
967 
968     return XXH_OK;
969 }
970 
XXH64_update(XXH64_state_t * state_in,const void * input,size_t len)971 XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void* input, size_t len)
972 {
973     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
974 
975     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
976         return XXH64_update_endian(state_in, input, len, XXH_littleEndian);
977     else
978         return XXH64_update_endian(state_in, input, len, XXH_bigEndian);
979 }
980 
XXH64_digest_endian(const XXH64_state_t * state,XXH_endianess endian)981 FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endianess endian)
982 {
983     U64 h64;
984 
985     if (state->total_len >= 32) {
986         U64 const v1 = state->v1;
987         U64 const v2 = state->v2;
988         U64 const v3 = state->v3;
989         U64 const v4 = state->v4;
990 
991         h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
992         h64 = XXH64_mergeRound(h64, v1);
993         h64 = XXH64_mergeRound(h64, v2);
994         h64 = XXH64_mergeRound(h64, v3);
995         h64 = XXH64_mergeRound(h64, v4);
996     } else {
997         h64  = state->v3 /*seed*/ + PRIME64_5;
998     }
999 
1000     h64 += (U64) state->total_len;
1001 
1002     return XXH64_finalize(h64, state->mem64, (size_t)state->total_len, endian, XXH_aligned);
1003 }
1004 
XXH64_digest(const XXH64_state_t * state_in)1005 XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state_in)
1006 {
1007     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
1008 
1009     if ((endian_detected==XXH_littleEndian) || XXH_FORCE_NATIVE_FORMAT)
1010         return XXH64_digest_endian(state_in, XXH_littleEndian);
1011     else
1012         return XXH64_digest_endian(state_in, XXH_bigEndian);
1013 }
1014 
1015 
1016 /*====== Canonical representation   ======*/
1017 
XXH64_canonicalFromHash(XXH64_canonical_t * dst,XXH64_hash_t hash)1018 XXH_PUBLIC_API void XXH64_canonicalFromHash(XXH64_canonical_t* dst, XXH64_hash_t hash)
1019 {
1020     XXH_STATIC_ASSERT(sizeof(XXH64_canonical_t) == sizeof(XXH64_hash_t));
1021     if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap64(hash);
1022     memcpy(dst, &hash, sizeof(*dst));
1023 }
1024 
XXH64_hashFromCanonical(const XXH64_canonical_t * src)1025 XXH_PUBLIC_API XXH64_hash_t XXH64_hashFromCanonical(const XXH64_canonical_t* src)
1026 {
1027     return XXH_readBE64(src);
1028 }
1029 
1030 #endif  /* XXH_NO_LONG_LONG */
1031