forked from facebook/rocksdb
-
Notifications
You must be signed in to change notification settings - Fork 0
/
xxhash.cc
1158 lines (981 loc) · 39.2 KB
/
xxhash.cc
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
/*
* xxHash - Fast Hash algorithm
* Copyright (C) 2012-2016, Yann Collet
*
* BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* You can contact the author at :
* - xxHash homepage: http://www.xxhash.com
* - xxHash source repository : https://github.com/Cyan4973/xxHash
*/
/* since xxhash.c can be included (via XXH_INLINE_ALL),
* it's good practice to protect it with guard
* in case of multiples inclusions */
#ifndef XXHASH_C_01393879
#define XXHASH_C_01393879
/* *************************************
* Tuning parameters
***************************************/
/*!XXH_FORCE_MEMORY_ACCESS :
* By default, access to unaligned memory is controlled by `memcpy()`, which is safe and portable.
* Unfortunately, on some target/compiler combinations, the generated assembly is sub-optimal.
* The below switch allow to select different access method for improved performance.
* Method 0 (default) : use `memcpy()`. Safe and portable.
* Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
* This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
* Method 2 : direct access. This method doesn't depend on compiler but violate C standard.
* It can generate buggy code on targets which do not support unaligned memory accesses.
* But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
* See http://stackoverflow.com/a/32095106/646947 for details.
* Prefer these methods in priority order (0 > 1 > 2)
*/
#ifndef XXH_FORCE_MEMORY_ACCESS /* can be defined externally, on command line for example */
# if !defined(__clang__) && defined(__GNUC__) && defined(__ARM_FEATURE_UNALIGNED) && defined(__ARM_ARCH) && (__ARM_ARCH == 6)
# define XXH_FORCE_MEMORY_ACCESS 2
# elif !defined(__clang__) && ((defined(__INTEL_COMPILER) && !defined(_WIN32)) || \
(defined(__GNUC__) && (defined(__ARM_ARCH) && __ARM_ARCH >= 7)))
# define XXH_FORCE_MEMORY_ACCESS 1
# endif
#endif
/*!XXH_ACCEPT_NULL_INPUT_POINTER :
* If input pointer is NULL, xxHash default behavior is to dereference it, triggering a segfault.
* When this macro is enabled, xxHash actively checks input for null pointer.
* It it is, result for null input pointers is the same as a null-length input.
*/
#ifndef XXH_ACCEPT_NULL_INPUT_POINTER /* can be defined externally */
# define XXH_ACCEPT_NULL_INPUT_POINTER 0
#endif
/*!XXH_FORCE_ALIGN_CHECK :
* This is a minor performance trick, only useful with lots of very small keys.
* It means : check for aligned/unaligned input.
* The check costs one initial branch per hash;
* set it to 0 when the input is guaranteed to be aligned,
* or when alignment doesn't matter for performance.
*/
#ifndef XXH_FORCE_ALIGN_CHECK /* can be defined externally */
# if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64)
# define XXH_FORCE_ALIGN_CHECK 0
# else
# define XXH_FORCE_ALIGN_CHECK 1
# endif
#endif
/*!XXH_REROLL:
* Whether to reroll XXH32_finalize, and XXH64_finalize,
* instead of using an unrolled jump table/if statement loop.
*
* This is automatically defined on -Os/-Oz on GCC and Clang. */
#ifndef XXH_REROLL
# if defined(__OPTIMIZE_SIZE__)
# define XXH_REROLL 1
# else
# define XXH_REROLL 0
# endif
#endif
/* *************************************
* Includes & Memory related functions
***************************************/
/*! Modify the local functions below should you wish to use some other memory routines
* for malloc(), free() */
#include <stdlib.h>
static void* XXH_malloc(size_t s) { return malloc(s); }
static void XXH_free (void* p) { free(p); }
/*! and for memcpy() */
#include <string.h>
static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
#include <limits.h> /* ULLONG_MAX */
#ifndef XXH_STATIC_LINKING_ONLY
#define XXH_STATIC_LINKING_ONLY
#endif
#include "xxhash.h"
/* BEGIN RocksDB customizations */
#include "port/lang.h" /* for FALLTHROUGH_INTENDED, inserted as appropriate */
/* END RocksDB customizations */
/* *************************************
* Compiler Specific Options
***************************************/
#ifdef _MSC_VER /* Visual Studio */
# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
# define XXH_FORCE_INLINE static __forceinline
# define XXH_NO_INLINE static __declspec(noinline)
#else
# if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */
# ifdef __GNUC__
# define XXH_FORCE_INLINE static inline __attribute__((always_inline))
# define XXH_NO_INLINE static __attribute__((noinline))
# else
# define XXH_FORCE_INLINE static inline
# define XXH_NO_INLINE static
# endif
# else
# define XXH_FORCE_INLINE static
# define XXH_NO_INLINE static
# endif /* __STDC_VERSION__ */
#endif
/* *************************************
* Debug
***************************************/
/* DEBUGLEVEL is expected to be defined externally,
* typically through compiler command line.
* Value must be a number. */
#ifndef DEBUGLEVEL
# define DEBUGLEVEL 0
#endif
#if (DEBUGLEVEL>=1)
# include <assert.h> /* note : can still be disabled with NDEBUG */
# define XXH_ASSERT(c) assert(c)
#else
# define XXH_ASSERT(c) ((void)0)
#endif
/* note : use after variable declarations */
#define XXH_STATIC_ASSERT(c) { enum { XXH_sa = 1/(int)(!!(c)) }; }
/* *************************************
* Basic Types
***************************************/
#if !defined (__VMS) \
&& (defined (__cplusplus) \
|| (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
# include <stdint.h>
typedef uint8_t xxh_u8;
#else
typedef unsigned char xxh_u8;
#endif
typedef XXH32_hash_t xxh_u32;
/* === Memory access === */
#if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2))
/* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */
static xxh_u32 XXH_read32(const void* memPtr) { return *(const xxh_u32*) memPtr; }
#elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1))
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { xxh_u32 u32; } __attribute__((packed)) unalign;
static xxh_u32 XXH_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
#else
/* portable and safe solution. Generally efficient.
* see : http://stackoverflow.com/a/32095106/646947
*/
static xxh_u32 XXH_read32(const void* memPtr)
{
xxh_u32 val;
memcpy(&val, memPtr, sizeof(val));
return val;
}
#endif /* XXH_FORCE_DIRECT_MEMORY_ACCESS */
/* === Endianess === */
/* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */
#ifndef XXH_CPU_LITTLE_ENDIAN
# if defined(_WIN32) /* Windows is always little endian */ \
|| defined(__LITTLE_ENDIAN__) \
|| (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__)
# define XXH_CPU_LITTLE_ENDIAN 1
# elif defined(__BIG_ENDIAN__) \
|| (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
# define XXH_CPU_LITTLE_ENDIAN 0
# else
static int XXH_isLittleEndian(void)
{
const union { xxh_u32 u; xxh_u8 c[4]; } one = { 1 }; /* don't use static : performance detrimental */
return one.c[0];
}
# define XXH_CPU_LITTLE_ENDIAN XXH_isLittleEndian()
# endif
#endif
/* ****************************************
* Compiler-specific Functions and Macros
******************************************/
#define XXH_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
#ifndef __has_builtin
# define __has_builtin(x) 0
#endif
#if !defined(NO_CLANG_BUILTIN) && __has_builtin(__builtin_rotateleft32) && __has_builtin(__builtin_rotateleft64)
# define XXH_rotl32 __builtin_rotateleft32
# define XXH_rotl64 __builtin_rotateleft64
/* Note : although _rotl exists for minGW (GCC under windows), performance seems poor */
#elif defined(_MSC_VER)
# define XXH_rotl32(x,r) _rotl(x,r)
# define XXH_rotl64(x,r) _rotl64(x,r)
#else
# define XXH_rotl32(x,r) (((x) << (r)) | ((x) >> (32 - (r))))
# define XXH_rotl64(x,r) (((x) << (r)) | ((x) >> (64 - (r))))
#endif
#if defined(_MSC_VER) /* Visual Studio */
# define XXH_swap32 _byteswap_ulong
#elif XXH_GCC_VERSION >= 403
# define XXH_swap32 __builtin_bswap32
#else
static xxh_u32 XXH_swap32 (xxh_u32 x)
{
return ((x << 24) & 0xff000000 ) |
((x << 8) & 0x00ff0000 ) |
((x >> 8) & 0x0000ff00 ) |
((x >> 24) & 0x000000ff );
}
#endif
/* ***************************
* Memory reads
*****************************/
typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment;
XXH_FORCE_INLINE xxh_u32 XXH_readLE32(const void* ptr)
{
return XXH_CPU_LITTLE_ENDIAN ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr));
}
static xxh_u32 XXH_readBE32(const void* ptr)
{
return XXH_CPU_LITTLE_ENDIAN ? XXH_swap32(XXH_read32(ptr)) : XXH_read32(ptr);
}
XXH_FORCE_INLINE xxh_u32
XXH_readLE32_align(const void* ptr, XXH_alignment align)
{
if (align==XXH_unaligned) {
return XXH_readLE32(ptr);
} else {
return XXH_CPU_LITTLE_ENDIAN ? *(const xxh_u32*)ptr : XXH_swap32(*(const xxh_u32*)ptr);
}
}
/* *************************************
* Misc
***************************************/
XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; }
/* *******************************************************************
* 32-bit hash functions
*********************************************************************/
static const xxh_u32 PRIME32_1 = 0x9E3779B1U; /* 0b10011110001101110111100110110001 */
static const xxh_u32 PRIME32_2 = 0x85EBCA77U; /* 0b10000101111010111100101001110111 */
static const xxh_u32 PRIME32_3 = 0xC2B2AE3DU; /* 0b11000010101100101010111000111101 */
static const xxh_u32 PRIME32_4 = 0x27D4EB2FU; /* 0b00100111110101001110101100101111 */
static const xxh_u32 PRIME32_5 = 0x165667B1U; /* 0b00010110010101100110011110110001 */
static xxh_u32 XXH32_round(xxh_u32 acc, xxh_u32 input)
{
acc += input * PRIME32_2;
acc = XXH_rotl32(acc, 13);
acc *= PRIME32_1;
#if defined(__GNUC__) && defined(__SSE4_1__) && !defined(XXH_ENABLE_AUTOVECTORIZE)
/* UGLY HACK:
* This inline assembly hack forces acc into a normal register. This is the
* only thing that prevents GCC and Clang from autovectorizing the XXH32 loop
* (pragmas and attributes don't work for some resason) without globally
* disabling SSE4.1.
*
* The reason we want to avoid vectorization is because despite working on
* 4 integers at a time, there are multiple factors slowing XXH32 down on
* SSE4:
* - There's a ridiculous amount of lag from pmulld (10 cycles of latency on newer chips!)
* making it slightly slower to multiply four integers at once compared to four
* integers independently. Even when pmulld was fastest, Sandy/Ivy Bridge, it is
* still not worth it to go into SSE just to multiply unless doing a long operation.
*
* - Four instructions are required to rotate,
* movqda tmp, v // not required with VEX encoding
* pslld tmp, 13 // tmp <<= 13
* psrld v, 19 // x >>= 19
* por v, tmp // x |= tmp
* compared to one for scalar:
* roll v, 13 // reliably fast across the board
* shldl v, v, 13 // Sandy Bridge and later prefer this for some reason
*
* - Instruction level parallelism is actually more beneficial here because the
* SIMD actually serializes this operation: While v1 is rotating, v2 can load data,
* while v3 can multiply. SSE forces them to operate together.
*
* How this hack works:
* __asm__("" // Declare an assembly block but don't declare any instructions
* : // However, as an Input/Output Operand,
* "+r" // constrain a read/write operand (+) as a general purpose register (r).
* (acc) // and set acc as the operand
* );
*
* Because of the 'r', the compiler has promised that seed will be in a
* general purpose register and the '+' says that it will be 'read/write',
* so it has to assume it has changed. It is like volatile without all the
* loads and stores.
*
* Since the argument has to be in a normal register (not an SSE register),
* each time XXH32_round is called, it is impossible to vectorize. */
__asm__("" : "+r" (acc));
#endif
return acc;
}
/* mix all bits */
static xxh_u32 XXH32_avalanche(xxh_u32 h32)
{
h32 ^= h32 >> 15;
h32 *= PRIME32_2;
h32 ^= h32 >> 13;
h32 *= PRIME32_3;
h32 ^= h32 >> 16;
return(h32);
}
#define XXH_get32bits(p) XXH_readLE32_align(p, align)
static xxh_u32
XXH32_finalize(xxh_u32 h32, const xxh_u8* ptr, size_t len, XXH_alignment align)
{
#define PROCESS1 \
h32 += (*ptr++) * PRIME32_5; \
h32 = XXH_rotl32(h32, 11) * PRIME32_1 ;
#define PROCESS4 \
h32 += XXH_get32bits(ptr) * PRIME32_3; \
ptr+=4; \
h32 = XXH_rotl32(h32, 17) * PRIME32_4 ;
/* Compact rerolled version */
if (XXH_REROLL) {
len &= 15;
while (len >= 4) {
PROCESS4;
len -= 4;
}
while (len > 0) {
PROCESS1;
--len;
}
return XXH32_avalanche(h32);
} else {
switch(len&15) /* or switch(bEnd - p) */ {
case 12: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 8: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 4: PROCESS4;
return XXH32_avalanche(h32);
case 13: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 9: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 5: PROCESS4;
PROCESS1;
return XXH32_avalanche(h32);
case 14: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 10: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 6: PROCESS4;
PROCESS1;
PROCESS1;
return XXH32_avalanche(h32);
case 15: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 11: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 7: PROCESS4;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 3: PROCESS1;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 2: PROCESS1;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 1: PROCESS1;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 0: return XXH32_avalanche(h32);
}
XXH_ASSERT(0);
return h32; /* reaching this point is deemed impossible */
}
}
XXH_FORCE_INLINE xxh_u32
XXH32_endian_align(const xxh_u8* input, size_t len, xxh_u32 seed, XXH_alignment align)
{
const xxh_u8* bEnd = input + len;
xxh_u32 h32;
#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
if (input==NULL) {
len=0;
bEnd=input=(const xxh_u8*)(size_t)16;
}
#endif
if (len>=16) {
const xxh_u8* const limit = bEnd - 15;
xxh_u32 v1 = seed + PRIME32_1 + PRIME32_2;
xxh_u32 v2 = seed + PRIME32_2;
xxh_u32 v3 = seed + 0;
xxh_u32 v4 = seed - PRIME32_1;
do {
v1 = XXH32_round(v1, XXH_get32bits(input)); input += 4;
v2 = XXH32_round(v2, XXH_get32bits(input)); input += 4;
v3 = XXH32_round(v3, XXH_get32bits(input)); input += 4;
v4 = XXH32_round(v4, XXH_get32bits(input)); input += 4;
} while (input < limit);
h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7)
+ XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
} else {
h32 = seed + PRIME32_5;
}
h32 += (xxh_u32)len;
return XXH32_finalize(h32, input, len&15, align);
}
XXH_PUBLIC_API XXH32_hash_t XXH32 (const void* input, size_t len, XXH32_hash_t seed)
{
#if 0
/* Simple version, good for code maintenance, but unfortunately slow for small inputs */
XXH32_state_t state;
XXH32_reset(&state, seed);
XXH32_update(&state, (const xxh_u8*)input, len);
return XXH32_digest(&state);
#else
if (XXH_FORCE_ALIGN_CHECK) {
if ((((size_t)input) & 3) == 0) { /* Input is 4-bytes aligned, leverage the speed benefit */
return XXH32_endian_align((const xxh_u8*)input, len, seed, XXH_aligned);
} }
return XXH32_endian_align((const xxh_u8*)input, len, seed, XXH_unaligned);
#endif
}
/*====== Hash streaming ======*/
XXH_PUBLIC_API XXH32_state_t* XXH32_createState(void)
{
return (XXH32_state_t*)XXH_malloc(sizeof(XXH32_state_t));
}
XXH_PUBLIC_API XXH_errorcode XXH32_freeState(XXH32_state_t* statePtr)
{
XXH_free(statePtr);
return XXH_OK;
}
XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* dstState, const XXH32_state_t* srcState)
{
memcpy(dstState, srcState, sizeof(*dstState));
}
XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, XXH32_hash_t seed)
{
XXH32_state_t state; /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
memset(&state, 0, sizeof(state));
state.v1 = seed + PRIME32_1 + PRIME32_2;
state.v2 = seed + PRIME32_2;
state.v3 = seed + 0;
state.v4 = seed - PRIME32_1;
/* do not write into reserved, planned to be removed in a future version */
memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved));
return XXH_OK;
}
XXH_PUBLIC_API XXH_errorcode
XXH32_update(XXH32_state_t* state, const void* input, size_t len)
{
if (input==NULL)
#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
return XXH_OK;
#else
return XXH_ERROR;
#endif
{ const xxh_u8* p = (const xxh_u8*)input;
const xxh_u8* const bEnd = p + len;
state->total_len_32 += (XXH32_hash_t)len;
state->large_len |= (XXH32_hash_t)((len>=16) | (state->total_len_32>=16));
if (state->memsize + len < 16) { /* fill in tmp buffer */
XXH_memcpy((xxh_u8*)(state->mem32) + state->memsize, input, len);
state->memsize += (XXH32_hash_t)len;
return XXH_OK;
}
if (state->memsize) { /* some data left from previous update */
XXH_memcpy((xxh_u8*)(state->mem32) + state->memsize, input, 16-state->memsize);
{ const xxh_u32* p32 = state->mem32;
state->v1 = XXH32_round(state->v1, XXH_readLE32(p32)); p32++;
state->v2 = XXH32_round(state->v2, XXH_readLE32(p32)); p32++;
state->v3 = XXH32_round(state->v3, XXH_readLE32(p32)); p32++;
state->v4 = XXH32_round(state->v4, XXH_readLE32(p32));
}
p += 16-state->memsize;
state->memsize = 0;
}
// uintptr_t casts added to avoid array-bounds error on
// some inlined calls
if ((uintptr_t)p <= (uintptr_t)bEnd - 16) {
const uintptr_t limit = (uintptr_t)bEnd - 16;
xxh_u32 v1 = state->v1;
xxh_u32 v2 = state->v2;
xxh_u32 v3 = state->v3;
xxh_u32 v4 = state->v4;
do {
v1 = XXH32_round(v1, XXH_readLE32(p)); p+=4;
v2 = XXH32_round(v2, XXH_readLE32(p)); p+=4;
v3 = XXH32_round(v3, XXH_readLE32(p)); p+=4;
v4 = XXH32_round(v4, XXH_readLE32(p)); p+=4;
} while ((uintptr_t)p <= limit);
state->v1 = v1;
state->v2 = v2;
state->v3 = v3;
state->v4 = v4;
}
if (p < bEnd) {
XXH_memcpy(state->mem32, p, (size_t)(bEnd-p));
state->memsize = (unsigned)(bEnd-p);
}
}
return XXH_OK;
}
XXH_PUBLIC_API XXH32_hash_t XXH32_digest (const XXH32_state_t* state)
{
xxh_u32 h32;
if (state->large_len) {
h32 = XXH_rotl32(state->v1, 1)
+ XXH_rotl32(state->v2, 7)
+ XXH_rotl32(state->v3, 12)
+ XXH_rotl32(state->v4, 18);
} else {
h32 = state->v3 /* == seed */ + PRIME32_5;
}
h32 += state->total_len_32;
return XXH32_finalize(h32, (const xxh_u8*)state->mem32, state->memsize, XXH_aligned);
}
/*====== Canonical representation ======*/
/*! Default XXH result types are basic unsigned 32 and 64 bits.
* The canonical representation follows human-readable write convention, aka big-endian (large digits first).
* These functions allow transformation of hash result into and from its canonical format.
* This way, hash values can be written into a file or buffer, remaining comparable across different systems.
*/
XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash)
{
XXH_STATIC_ASSERT(sizeof(XXH32_canonical_t) == sizeof(XXH32_hash_t));
if (XXH_CPU_LITTLE_ENDIAN) hash = XXH_swap32(hash);
memcpy(dst, &hash, sizeof(*dst));
}
XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src)
{
return XXH_readBE32(src);
}
#ifndef XXH_NO_LONG_LONG
/* *******************************************************************
* 64-bit hash functions
*********************************************************************/
/*====== Memory access ======*/
typedef XXH64_hash_t xxh_u64;
/*! XXH_REROLL_XXH64:
* Whether to reroll the XXH64_finalize() loop.
*
* Just like XXH32, we can unroll the XXH64_finalize() loop. This can be a performance gain
* on 64-bit hosts, as only one jump is required.
*
* However, on 32-bit hosts, because arithmetic needs to be done with two 32-bit registers,
* and 64-bit arithmetic needs to be simulated, it isn't beneficial to unroll. The code becomes
* ridiculously large (the largest function in the binary on i386!), and rerolling it saves
* anywhere from 3kB to 20kB. It is also slightly faster because it fits into cache better
* and is more likely to be inlined by the compiler.
*
* If XXH_REROLL is defined, this is ignored and the loop is always rerolled. */
#ifndef XXH_REROLL_XXH64
# if (defined(__ILP32__) || defined(_ILP32)) /* ILP32 is often defined on 32-bit GCC family */ \
|| !(defined(__x86_64__) || defined(_M_X64) || defined(_M_AMD64) /* x86-64 */ \
|| defined(_M_ARM64) || defined(__aarch64__) || defined(__arm64__) /* aarch64 */ \
|| defined(__PPC64__) || defined(__PPC64LE__) || defined(__ppc64__) || defined(__powerpc64__) /* ppc64 */ \
|| defined(__mips64__) || defined(__mips64)) /* mips64 */ \
|| (!defined(SIZE_MAX) || SIZE_MAX < ULLONG_MAX) /* check limits */
# define XXH_REROLL_XXH64 1
# else
# define XXH_REROLL_XXH64 0
# endif
#endif /* !defined(XXH_REROLL_XXH64) */
#if (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==2))
/* Force direct memory access. Only works on CPU which support unaligned memory access in hardware */
static xxh_u64 XXH_read64(const void* memPtr) { return *(const xxh_u64*) memPtr; }
#elif (defined(XXH_FORCE_MEMORY_ACCESS) && (XXH_FORCE_MEMORY_ACCESS==1))
/* __pack instructions are safer, but compiler specific, hence potentially problematic for some compilers */
/* currently only defined for gcc and icc */
typedef union { xxh_u32 u32; xxh_u64 u64; } __attribute__((packed)) unalign64;
static xxh_u64 XXH_read64(const void* ptr) { return ((const unalign64*)ptr)->u64; }
#else
/* portable and safe solution. Generally efficient.
* see : http://stackoverflow.com/a/32095106/646947
*/
static xxh_u64 XXH_read64(const void* memPtr)
{
xxh_u64 val;
memcpy(&val, memPtr, sizeof(val));
return val;
}
#endif /* XXH_FORCE_DIRECT_MEMORY_ACCESS */
#if defined(_MSC_VER) /* Visual Studio */
# define XXH_swap64 _byteswap_uint64
#elif XXH_GCC_VERSION >= 403
# define XXH_swap64 __builtin_bswap64
#else
static xxh_u64 XXH_swap64 (xxh_u64 x)
{
return ((x << 56) & 0xff00000000000000ULL) |
((x << 40) & 0x00ff000000000000ULL) |
((x << 24) & 0x0000ff0000000000ULL) |
((x << 8) & 0x000000ff00000000ULL) |
((x >> 8) & 0x00000000ff000000ULL) |
((x >> 24) & 0x0000000000ff0000ULL) |
((x >> 40) & 0x000000000000ff00ULL) |
((x >> 56) & 0x00000000000000ffULL);
}
#endif
XXH_FORCE_INLINE xxh_u64 XXH_readLE64(const void* ptr)
{
return XXH_CPU_LITTLE_ENDIAN ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr));
}
static xxh_u64 XXH_readBE64(const void* ptr)
{
return XXH_CPU_LITTLE_ENDIAN ? XXH_swap64(XXH_read64(ptr)) : XXH_read64(ptr);
}
XXH_FORCE_INLINE xxh_u64
XXH_readLE64_align(const void* ptr, XXH_alignment align)
{
if (align==XXH_unaligned)
return XXH_readLE64(ptr);
else
return XXH_CPU_LITTLE_ENDIAN ? *(const xxh_u64*)ptr : XXH_swap64(*(const xxh_u64*)ptr);
}
/*====== xxh64 ======*/
static const xxh_u64 PRIME64_1 = 0x9E3779B185EBCA87ULL; /* 0b1001111000110111011110011011000110000101111010111100101010000111 */
static const xxh_u64 PRIME64_2 = 0xC2B2AE3D27D4EB4FULL; /* 0b1100001010110010101011100011110100100111110101001110101101001111 */
static const xxh_u64 PRIME64_3 = 0x165667B19E3779F9ULL; /* 0b0001011001010110011001111011000110011110001101110111100111111001 */
static const xxh_u64 PRIME64_4 = 0x85EBCA77C2B2AE63ULL; /* 0b1000010111101011110010100111011111000010101100101010111001100011 */
static const xxh_u64 PRIME64_5 = 0x27D4EB2F165667C5ULL; /* 0b0010011111010100111010110010111100010110010101100110011111000101 */
static xxh_u64 XXH64_round(xxh_u64 acc, xxh_u64 input)
{
acc += input * PRIME64_2;
acc = XXH_rotl64(acc, 31);
acc *= PRIME64_1;
return acc;
}
static xxh_u64 XXH64_mergeRound(xxh_u64 acc, xxh_u64 val)
{
val = XXH64_round(0, val);
acc ^= val;
acc = acc * PRIME64_1 + PRIME64_4;
return acc;
}
static xxh_u64 XXH64_avalanche(xxh_u64 h64)
{
h64 ^= h64 >> 33;
h64 *= PRIME64_2;
h64 ^= h64 >> 29;
h64 *= PRIME64_3;
h64 ^= h64 >> 32;
return h64;
}
#define XXH_get64bits(p) XXH_readLE64_align(p, align)
static xxh_u64
XXH64_finalize(xxh_u64 h64, const xxh_u8* ptr, size_t len, XXH_alignment align)
{
#define PROCESS1_64 \
h64 ^= (*ptr++) * PRIME64_5; \
h64 = XXH_rotl64(h64, 11) * PRIME64_1;
#define PROCESS4_64 \
h64 ^= (xxh_u64)(XXH_get32bits(ptr)) * PRIME64_1; \
ptr+=4; \
h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
#define PROCESS8_64 { \
xxh_u64 const k1 = XXH64_round(0, XXH_get64bits(ptr)); \
ptr+=8; \
h64 ^= k1; \
h64 = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; \
}
/* Rerolled version for 32-bit targets is faster and much smaller. */
if (XXH_REROLL || XXH_REROLL_XXH64) {
len &= 31;
while (len >= 8) {
PROCESS8_64;
len -= 8;
}
if (len >= 4) {
PROCESS4_64;
len -= 4;
}
while (len > 0) {
PROCESS1_64;
--len;
}
return XXH64_avalanche(h64);
} else {
switch(len & 31) {
case 24: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 16: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 8: PROCESS8_64;
return XXH64_avalanche(h64);
case 28: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 20: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 12: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 4: PROCESS4_64;
return XXH64_avalanche(h64);
case 25: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 17: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 9: PROCESS8_64;
PROCESS1_64;
return XXH64_avalanche(h64);
case 29: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 21: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 13: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 5: PROCESS4_64;
PROCESS1_64;
return XXH64_avalanche(h64);
case 26: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 18: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 10: PROCESS8_64;
PROCESS1_64;
PROCESS1_64;
return XXH64_avalanche(h64);
case 30: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 22: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 14: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 6: PROCESS4_64;
PROCESS1_64;
PROCESS1_64;
return XXH64_avalanche(h64);
case 27: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 19: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 11: PROCESS8_64;
PROCESS1_64;
PROCESS1_64;
PROCESS1_64;
return XXH64_avalanche(h64);
case 31: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 23: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 15: PROCESS8_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 7: PROCESS4_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 3: PROCESS1_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 2: PROCESS1_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 1: PROCESS1_64;
FALLTHROUGH_INTENDED;
/* fallthrough */
case 0: return XXH64_avalanche(h64);
}
}
/* impossible to reach */
XXH_ASSERT(0);
return 0; /* unreachable, but some compilers complain without it */
}
XXH_FORCE_INLINE xxh_u64
XXH64_endian_align(const xxh_u8* input, size_t len, xxh_u64 seed, XXH_alignment align)
{
const xxh_u8* bEnd = input + len;
xxh_u64 h64;
#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
if (input==NULL) {
len=0;
bEnd=input=(const xxh_u8*)(size_t)32;
}
#endif
if (len>=32) {
const xxh_u8* const limit = bEnd - 32;
xxh_u64 v1 = seed + PRIME64_1 + PRIME64_2;
xxh_u64 v2 = seed + PRIME64_2;
xxh_u64 v3 = seed + 0;
xxh_u64 v4 = seed - PRIME64_1;
do {
v1 = XXH64_round(v1, XXH_get64bits(input)); input+=8;
v2 = XXH64_round(v2, XXH_get64bits(input)); input+=8;
v3 = XXH64_round(v3, XXH_get64bits(input)); input+=8;
v4 = XXH64_round(v4, XXH_get64bits(input)); input+=8;
} while (input<=limit);
h64 = XXH_rotl64(v1, 1) + XXH_rotl64(v2, 7) + XXH_rotl64(v3, 12) + XXH_rotl64(v4, 18);
h64 = XXH64_mergeRound(h64, v1);
h64 = XXH64_mergeRound(h64, v2);
h64 = XXH64_mergeRound(h64, v3);
h64 = XXH64_mergeRound(h64, v4);
} else {
h64 = seed + PRIME64_5;
}
h64 += (xxh_u64) len;
return XXH64_finalize(h64, input, len, align);
}
XXH_PUBLIC_API XXH64_hash_t XXH64 (const void* input, size_t len, XXH64_hash_t seed)
{
#if 0
/* Simple version, good for code maintenance, but unfortunately slow for small inputs */
XXH64_state_t state;
XXH64_reset(&state, seed);
XXH64_update(&state, (const xxh_u8*)input, len);