forked from demerphq/smhasher
-
Notifications
You must be signed in to change notification settings - Fork 0
/
farmhash.cc
11870 lines (11390 loc) · 277 KB
/
farmhash.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) 2014 Google, Inc.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
//
// FarmHash, by Geoff Pike
#include "farmhash.h"
// FARMHASH ASSUMPTIONS: Modify as needed, or use -DFARMHASH_ASSUME_SSE42 etc.
// Note that if you use -DFARMHASH_ASSUME_SSE42 you likely need -msse42
// (or its equivalent for your compiler); if you use -DFARMHASH_ASSUME_AESNI
// you likely need -maes (or its equivalent for your compiler).
#ifdef FARMHASH_ASSUME_SSSE3
#undef FARMHASH_ASSUME_SSSE3
#define FARMHASH_ASSUME_SSSE3
#endif
#ifdef FARMHASH_ASSUME_SSE41
#undef FARMHASH_ASSUME_SSE41
#define FARMHASH_ASSUME_SSE41
#endif
#ifdef FARMHASH_ASSUME_SSE42
#undef FARMHASH_ASSUME_SSE42
#define FARMHASH_ASSUME_SSE42
#endif
#ifdef FARMHASH_ASSUME_AESNI
#undef FARMHASH_ASSUME_AESNI
#define FARMHASH_ASSUME_AESNI
#endif
#ifdef FARMHASH_ASSUME_AVX
#undef FARMHASH_ASSUME_AVX
#define FARMHASH_ASSUME_AVX
#endif
#if !defined(FARMHASH_CAN_USE_CXX11) && defined(LANG_CXX11)
#define FARMHASH_CAN_USE_CXX11 1
#else
#undef FARMHASH_CAN_USE_CXX11
#define FARMHASH_CAN_USE_CXX11 0
#endif
// FARMHASH PORTABILITY LAYER: Runtime error if misconfigured
#ifndef FARMHASH_DIE_IF_MISCONFIGURED
#define FARMHASH_DIE_IF_MISCONFIGURED do { *(char*)(len % 17) = 0; } while (0)
#endif
// FARMHASH PORTABILITY LAYER: "static inline" or similar
#ifndef STATIC_INLINE
#define STATIC_INLINE static inline
#endif
// FARMHASH PORTABILITY LAYER: LIKELY and UNLIKELY
#if !defined(LIKELY)
#if defined(FARMHASH_NO_BUILTIN_EXPECT) || !defined(HAVE_BUILTIN_EXPECT)
#define LIKELY(x) (x)
#else
#define LIKELY(x) (__builtin_expect(!!(x), 1))
#endif
#endif
#undef UNLIKELY
#define UNLIKELY(x) !LIKELY(!(x))
// FARMHASH PORTABILITY LAYER: endianness and byteswapping functions
#ifdef WORDS_BIGENDIAN
#undef FARMHASH_BIG_ENDIAN
#define FARMHASH_BIG_ENDIAN 1
#endif
#if defined(FARMHASH_LITTLE_ENDIAN) && defined(FARMHASH_BIG_ENDIAN)
#error
#endif
#if !defined(FARMHASH_LITTLE_ENDIAN) && !defined(FARMHASH_BIG_ENDIAN)
#define FARMHASH_UNKNOWN_ENDIAN 1
#endif
#if !defined(bswap_32) || !defined(bswap_64)
#undef bswap_32
#undef bswap_64
#if defined(HAVE_BUILTIN_BSWAP) || defined(__clang__) || \
(defined(__GNUC__) && ((__GNUC__ == 4 && __GNUC_MINOR__ >= 8) || \
__GNUC__ >= 5))
// Easy case for bswap: no header file needed.
#define bswap_32(x) __builtin_bswap32(x)
#define bswap_64(x) __builtin_bswap64(x)
#endif
#endif
#if defined(FARMHASH_UNKNOWN_ENDIAN) || !defined(bswap_64)
#ifdef _MSC_VER
#undef bswap_32
#undef bswap_64
#define bswap_32(x) _byteswap_ulong(x)
#define bswap_64(x) _byteswap_uint64(x)
#elif defined(__APPLE__)
// Mac OS X / Darwin features
#include <libkern/OSByteOrder.h>
#undef bswap_32
#undef bswap_64
#define bswap_32(x) OSSwapInt32(x)
#define bswap_64(x) OSSwapInt64(x)
#elif defined(__sun) || defined(sun)
#include <sys/byteorder.h>
#undef bswap_32
#undef bswap_64
#define bswap_32(x) BSWAP_32(x)
#define bswap_64(x) BSWAP_64(x)
#elif defined(__FreeBSD__)
#include <sys/endian.h>
#undef bswap_32
#undef bswap_64
#define bswap_32(x) bswap32(x)
#define bswap_64(x) bswap64(x)
#elif defined(__OpenBSD__)
#include <sys/types.h>
#undef bswap_32
#undef bswap_64
#define bswap_32(x) swap32(x)
#define bswap_64(x) swap64(x)
#elif defined(__NetBSD__)
#include <sys/types.h>
#include <machine/bswap.h>
#if defined(__BSWAP_RENAME) && !defined(__bswap_32)
#undef bswap_32
#undef bswap_64
#define bswap_32(x) bswap32(x)
#define bswap_64(x) bswap64(x)
#endif
#else
#undef bswap_32
#undef bswap_64
#include <byteswap.h>
#endif
#ifdef WORDS_BIGENDIAN
#define FARMHASH_BIG_ENDIAN 1
#endif
#endif
#ifdef FARMHASH_BIG_ENDIAN
#define uint32_in_expected_order(x) (bswap_32(x))
#define uint64_in_expected_order(x) (bswap_64(x))
#else
#define uint32_in_expected_order(x) (x)
#define uint64_in_expected_order(x) (x)
#endif
namespace NAMESPACE_FOR_HASH_FUNCTIONS {
STATIC_INLINE uint64_t Fetch64(const char *p) {
uint64_t result;
memcpy(&result, p, sizeof(result));
return uint64_in_expected_order(result);
}
STATIC_INLINE uint32_t Fetch32(const char *p) {
uint32_t result;
memcpy(&result, p, sizeof(result));
return uint32_in_expected_order(result);
}
STATIC_INLINE uint32_t Bswap32(uint32_t val) { return bswap_32(val); }
STATIC_INLINE uint64_t Bswap64(uint64_t val) { return bswap_64(val); }
// FARMHASH PORTABILITY LAYER: bitwise rot
STATIC_INLINE uint32_t BasicRotate32(uint32_t val, int shift) {
// Avoid shifting by 32: doing so yields an undefined result.
return shift == 0 ? val : ((val >> shift) | (val << (32 - shift)));
}
STATIC_INLINE uint64_t BasicRotate64(uint64_t val, int shift) {
// Avoid shifting by 64: doing so yields an undefined result.
return shift == 0 ? val : ((val >> shift) | (val << (64 - shift)));
}
#if defined(_MSC_VER) && defined(FARMHASH_ROTR)
STATIC_INLINE uint32_t Rotate32(uint32_t val, int shift) {
return sizeof(unsigned long) == sizeof(val) ?
_lrotr(val, shift) :
BasicRotate32(val, shift);
}
STATIC_INLINE uint64_t Rotate64(uint64_t val, int shift) {
return sizeof(unsigned long) == sizeof(val) ?
_lrotr(val, shift) :
BasicRotate64(val, shift);
}
#else
STATIC_INLINE uint32_t Rotate32(uint32_t val, int shift) {
return BasicRotate32(val, shift);
}
STATIC_INLINE uint64_t Rotate64(uint64_t val, int shift) {
return BasicRotate64(val, shift);
}
#endif
} // namespace NAMESPACE_FOR_HASH_FUNCTIONS
// FARMHASH PORTABILITY LAYER: debug mode or max speed?
// One may use -DFARMHASH_DEBUG=1 or -DFARMHASH_DEBUG=0 to force the issue.
#if !defined(FARMHASH_DEBUG) && (!defined(NDEBUG) || defined(_DEBUG))
#define FARMHASH_DEBUG 1
#endif
#undef debug_mode
#if FARMHASH_DEBUG
#define debug_mode 1
#else
#define debug_mode 0
#endif
// PLATFORM-SPECIFIC FUNCTIONS AND MACROS
// TODO: See https://stackoverflow.com/questions/11228855/header-files-for-simd-intrinsics
#undef x86_64
#if defined (__x86_64) || defined (__x86_64__)
#define x86_64 1
#else
#define x86_64 0
#endif
#undef x86
#if defined(__i386__) || defined(__i386) || defined(__X86__)
#define x86 1
#else
#define x86 x86_64
#endif
#if !defined(is_64bit)
#define is_64bit (x86_64 || (sizeof(void*) == 8))
#endif
#undef can_use_ssse3
#if defined(__SSSE3__) || defined(FARMHASH_ASSUME_SSSE3)
# ifdef __GNUC__
# define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
# if GCC_VERSION > 403
# include <immintrin.h>
# define can_use_ssse3 1
# else
# define can_use_ssse3 0
# endif
# else
# include <immintrin.h>
# define can_use_ssse3 1
# endif
// Now we can use _mm_hsub_epi16 and so on.
#else
# define can_use_ssse3 0
#endif
#undef can_use_sse41
#if can_use_ssse3 && (defined(__SSE4_1__) || defined(FARMHASH_ASSUME_SSE41))
#include <immintrin.h>
#define can_use_sse41 1
// Now we can use _mm_insert_epi64 and so on.
#else
#define can_use_sse41 0
#endif
#undef can_use_sse42
#if defined(__SSE4_2__) || defined(FARMHASH_ASSUME_SSE42)
#include <nmmintrin.h>
#define can_use_sse42 1
// Now we can use _mm_crc32_u{32,16,8}. And on 64-bit platforms, _mm_crc32_u64.
#else
#define can_use_sse42 0
#endif
#undef can_use_aesni
#if defined(__AES__) || defined(FARMHASH_ASSUME_AESNI)
#include <wmmintrin.h>
#define can_use_aesni 1
// Now we can use _mm_aesimc_si128 and so on.
#else
#define can_use_aesni 0
#endif
#undef can_use_avx
#if defined(__AVX__) || defined(FARMHASH_ASSUME_AVX)
#include <immintrin.h>
#define can_use_avx 1
#else
#define can_use_avx 0
#endif
#if can_use_ssse3 || can_use_sse41 || can_use_sse42 || can_use_aesni || can_use_avx
STATIC_INLINE __m128i Fetch128(const char* s) {
return _mm_loadu_si128(reinterpret_cast<const __m128i*>(s));
}
#endif
// Building blocks for hash functions
// std::swap() was in <algorithm> but is in <utility> from C++11 on.
#if !FARMHASH_CAN_USE_CXX11
#include <algorithm>
#endif
#undef PERMUTE3
#define PERMUTE3(a, b, c) do { std::swap(a, b); std::swap(a, c); } while (0)
namespace NAMESPACE_FOR_HASH_FUNCTIONS {
// Some primes between 2^63 and 2^64 for various uses.
static const uint64_t k0 = 0xc3a5c85c97cb3127ULL;
static const uint64_t k1 = 0xb492b66fbe98f273ULL;
static const uint64_t k2 = 0x9ae16a3b2f90404fULL;
// Magic numbers for 32-bit hashing. Copied from Murmur3.
static const uint32_t c1 = 0xcc9e2d51;
static const uint32_t c2 = 0x1b873593;
// A 32-bit to 32-bit integer hash copied from Murmur3.
STATIC_INLINE uint32_t fmix(uint32_t h)
{
h ^= h >> 16;
h *= 0x85ebca6b;
h ^= h >> 13;
h *= 0xc2b2ae35;
h ^= h >> 16;
return h;
}
STATIC_INLINE uint32_t Mur(uint32_t a, uint32_t h) {
// Helper from Murmur3 for combining two 32-bit values.
a *= c1;
a = Rotate32(a, 17);
a *= c2;
h ^= a;
h = Rotate32(h, 19);
return h * 5 + 0xe6546b64;
}
template <typename T> STATIC_INLINE T DebugTweak(T x) {
if (debug_mode) {
if (sizeof(x) == 4) {
x = ~Bswap32(x * c1);
} else {
x = ~Bswap64(x * k1);
}
}
return x;
}
template <> uint128_t DebugTweak(uint128_t x) {
if (debug_mode) {
uint64_t y = DebugTweak(Uint128Low64(x));
uint64_t z = DebugTweak(Uint128High64(x));
y += z;
z += y;
x = Uint128(y, z * k1);
}
return x;
}
} // namespace NAMESPACE_FOR_HASH_FUNCTIONS
using namespace std;
using namespace NAMESPACE_FOR_HASH_FUNCTIONS;
namespace farmhashna {
#undef Fetch
#define Fetch Fetch64
#undef Rotate
#define Rotate Rotate64
#undef Bswap
#define Bswap Bswap64
STATIC_INLINE uint64_t ShiftMix(uint64_t val) {
return val ^ (val >> 47);
}
STATIC_INLINE uint64_t HashLen16(uint64_t u, uint64_t v) {
return Hash128to64(Uint128(u, v));
}
STATIC_INLINE uint64_t HashLen16(uint64_t u, uint64_t v, uint64_t mul) {
// Murmur-inspired hashing.
uint64_t a = (u ^ v) * mul;
a ^= (a >> 47);
uint64_t b = (v ^ a) * mul;
b ^= (b >> 47);
b *= mul;
return b;
}
STATIC_INLINE uint64_t HashLen0to16(const char *s, size_t len) {
if (len >= 8) {
uint64_t mul = k2 + len * 2;
uint64_t a = Fetch(s) + k2;
uint64_t b = Fetch(s + len - 8);
uint64_t c = Rotate(b, 37) * mul + a;
uint64_t d = (Rotate(a, 25) + b) * mul;
return HashLen16(c, d, mul);
}
if (len >= 4) {
uint64_t mul = k2 + len * 2;
uint64_t a = Fetch32(s);
return HashLen16(len + (a << 3), Fetch32(s + len - 4), mul);
}
if (len > 0) {
uint8_t a = s[0];
uint8_t b = s[len >> 1];
uint8_t c = s[len - 1];
uint32_t y = static_cast<uint32_t>(a) + (static_cast<uint32_t>(b) << 8);
uint32_t z = len + (static_cast<uint32_t>(c) << 2);
return ShiftMix(y * k2 ^ z * k0) * k2;
}
return k2;
}
// This probably works well for 16-byte strings as well, but it may be overkill
// in that case.
STATIC_INLINE uint64_t HashLen17to32(const char *s, size_t len) {
uint64_t mul = k2 + len * 2;
uint64_t a = Fetch(s) * k1;
uint64_t b = Fetch(s + 8);
uint64_t c = Fetch(s + len - 8) * mul;
uint64_t d = Fetch(s + len - 16) * k2;
return HashLen16(Rotate(a + b, 43) + Rotate(c, 30) + d,
a + Rotate(b + k2, 18) + c, mul);
}
// Return a 16-byte hash for 48 bytes. Quick and dirty.
// Callers do best to use "random-looking" values for a and b.
STATIC_INLINE pair<uint64_t, uint64_t> WeakHashLen32WithSeeds(
uint64_t w, uint64_t x, uint64_t y, uint64_t z, uint64_t a, uint64_t b) {
a += w;
b = Rotate(b + a + z, 21);
uint64_t c = a;
a += x;
a += y;
b += Rotate(a, 44);
return make_pair(a + z, b + c);
}
// Return a 16-byte hash for s[0] ... s[31], a, and b. Quick and dirty.
STATIC_INLINE pair<uint64_t, uint64_t> WeakHashLen32WithSeeds(
const char* s, uint64_t a, uint64_t b) {
return WeakHashLen32WithSeeds(Fetch(s),
Fetch(s + 8),
Fetch(s + 16),
Fetch(s + 24),
a,
b);
}
// Return an 8-byte hash for 33 to 64 bytes.
STATIC_INLINE uint64_t HashLen33to64(const char *s, size_t len) {
uint64_t mul = k2 + len * 2;
uint64_t a = Fetch(s) * k2;
uint64_t b = Fetch(s + 8);
uint64_t c = Fetch(s + len - 8) * mul;
uint64_t d = Fetch(s + len - 16) * k2;
uint64_t y = Rotate(a + b, 43) + Rotate(c, 30) + d;
uint64_t z = HashLen16(y, a + Rotate(b + k2, 18) + c, mul);
uint64_t e = Fetch(s + 16) * mul;
uint64_t f = Fetch(s + 24);
uint64_t g = (y + Fetch(s + len - 32)) * mul;
uint64_t h = (z + Fetch(s + len - 24)) * mul;
return HashLen16(Rotate(e + f, 43) + Rotate(g, 30) + h,
e + Rotate(f + a, 18) + g, mul);
}
uint64_t Hash64(const char *s, size_t len) {
const uint64_t seed = 81;
if (len <= 32) {
if (len <= 16) {
return HashLen0to16(s, len);
} else {
return HashLen17to32(s, len);
}
} else if (len <= 64) {
return HashLen33to64(s, len);
}
// For strings over 64 bytes we loop. Internal state consists of
// 56 bytes: v, w, x, y, and z.
uint64_t x = seed;
uint64_t y = seed * k1 + 113;
uint64_t z = ShiftMix(y * k2 + 113) * k2;
pair<uint64_t, uint64_t> v = make_pair(0, 0);
pair<uint64_t, uint64_t> w = make_pair(0, 0);
x = x * k2 + Fetch(s);
// Set end so that after the loop we have 1 to 64 bytes left to process.
const char* end = s + ((len - 1) / 64) * 64;
const char* last64 = end + ((len - 1) & 63) - 63;
assert(s + len - 64 == last64);
do {
x = Rotate(x + y + v.first + Fetch(s + 8), 37) * k1;
y = Rotate(y + v.second + Fetch(s + 48), 42) * k1;
x ^= w.second;
y += v.first + Fetch(s + 40);
z = Rotate(z + w.first, 33) * k1;
v = WeakHashLen32WithSeeds(s, v.second * k1, x + w.first);
w = WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch(s + 16));
std::swap(z, x);
s += 64;
} while (s != end);
uint64_t mul = k1 + ((z & 0xff) << 1);
// Make s point to the last 64 bytes of input.
s = last64;
w.first += ((len - 1) & 63);
v.first += w.first;
w.first += v.first;
x = Rotate(x + y + v.first + Fetch(s + 8), 37) * mul;
y = Rotate(y + v.second + Fetch(s + 48), 42) * mul;
x ^= w.second * 9;
y += v.first * 9 + Fetch(s + 40);
z = Rotate(z + w.first, 33) * mul;
v = WeakHashLen32WithSeeds(s, v.second * mul, x + w.first);
w = WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch(s + 16));
std::swap(z, x);
return HashLen16(HashLen16(v.first, w.first, mul) + ShiftMix(y) * k0 + z,
HashLen16(v.second, w.second, mul) + x,
mul);
}
uint64_t Hash64WithSeeds(const char *s, size_t len, uint64_t seed0, uint64_t seed1);
uint64_t Hash64WithSeed(const char *s, size_t len, uint64_t seed) {
return Hash64WithSeeds(s, len, k2, seed);
}
uint64_t Hash64WithSeeds(const char *s, size_t len, uint64_t seed0, uint64_t seed1) {
return HashLen16(Hash64(s, len) - seed0, seed1);
}
} // namespace farmhashna
namespace farmhashuo {
#undef Fetch
#define Fetch Fetch64
#undef Rotate
#define Rotate Rotate64
STATIC_INLINE uint64_t H(uint64_t x, uint64_t y, uint64_t mul, int r) {
uint64_t a = (x ^ y) * mul;
a ^= (a >> 47);
uint64_t b = (y ^ a) * mul;
return Rotate(b, r) * mul;
}
uint64_t Hash64WithSeeds(const char *s, size_t len,
uint64_t seed0, uint64_t seed1) {
if (len <= 64) {
return farmhashna::Hash64WithSeeds(s, len, seed0, seed1);
}
// For strings over 64 bytes we loop. Internal state consists of
// 64 bytes: u, v, w, x, y, and z.
uint64_t x = seed0;
uint64_t y = seed1 * k2 + 113;
uint64_t z = farmhashna::ShiftMix(y * k2) * k2;
pair<uint64_t, uint64_t> v = make_pair(seed0, seed1);
pair<uint64_t, uint64_t> w = make_pair(0, 0);
uint64_t u = x - z;
x *= k2;
uint64_t mul = k2 + (u & 0x82);
// Set end so that after the loop we have 1 to 64 bytes left to process.
const char* end = s + ((len - 1) / 64) * 64;
const char* last64 = end + ((len - 1) & 63) - 63;
assert(s + len - 64 == last64);
do {
uint64_t a0 = Fetch(s);
uint64_t a1 = Fetch(s + 8);
uint64_t a2 = Fetch(s + 16);
uint64_t a3 = Fetch(s + 24);
uint64_t a4 = Fetch(s + 32);
uint64_t a5 = Fetch(s + 40);
uint64_t a6 = Fetch(s + 48);
uint64_t a7 = Fetch(s + 56);
x += a0 + a1;
y += a2;
z += a3;
v.first += a4;
v.second += a5 + a1;
w.first += a6;
w.second += a7;
x = Rotate(x, 26);
x *= 9;
y = Rotate(y, 29);
z *= mul;
v.first = Rotate(v.first, 33);
v.second = Rotate(v.second, 30);
w.first ^= x;
w.first *= 9;
z = Rotate(z, 32);
z += w.second;
w.second += z;
z *= 9;
std::swap(u, y);
z += a0 + a6;
v.first += a2;
v.second += a3;
w.first += a4;
w.second += a5 + a6;
x += a1;
y += a7;
y += v.first;
v.first += x - y;
v.second += w.first;
w.first += v.second;
w.second += x - y;
x += w.second;
w.second = Rotate(w.second, 34);
std::swap(u, z);
s += 64;
} while (s != end);
// Make s point to the last 64 bytes of input.
s = last64;
u *= 9;
v.second = Rotate(v.second, 28);
v.first = Rotate(v.first, 20);
w.first += ((len - 1) & 63);
u += y;
y += u;
x = Rotate(y - x + v.first + Fetch(s + 8), 37) * mul;
y = Rotate(y ^ v.second ^ Fetch(s + 48), 42) * mul;
x ^= w.second * 9;
y += v.first + Fetch(s + 40);
z = Rotate(z + w.first, 33) * mul;
v = farmhashna::WeakHashLen32WithSeeds(s, v.second * mul, x + w.first);
w = farmhashna::WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch(s + 16));
return H(farmhashna::HashLen16(v.first + x, w.first ^ y, mul) + z - u,
H(v.second + y, w.second + z, k2, 30) ^ x,
k2,
31);
}
uint64_t Hash64WithSeed(const char *s, size_t len, uint64_t seed) {
return len <= 64 ? farmhashna::Hash64WithSeed(s, len, seed) :
Hash64WithSeeds(s, len, 0, seed);
}
uint64_t Hash64(const char *s, size_t len) {
return len <= 64 ? farmhashna::Hash64(s, len) :
Hash64WithSeeds(s, len, 81, 0);
}
} // namespace farmhashuo
namespace farmhashxo {
#undef Fetch
#define Fetch Fetch64
#undef Rotate
#define Rotate Rotate64
STATIC_INLINE uint64_t H32(const char *s, size_t len, uint64_t mul,
uint64_t seed0 = 0, uint64_t seed1 = 0) {
uint64_t a = Fetch(s) * k1;
uint64_t b = Fetch(s + 8);
uint64_t c = Fetch(s + len - 8) * mul;
uint64_t d = Fetch(s + len - 16) * k2;
uint64_t u = Rotate(a + b, 43) + Rotate(c, 30) + d + seed0;
uint64_t v = a + Rotate(b + k2, 18) + c + seed1;
a = farmhashna::ShiftMix((u ^ v) * mul);
b = farmhashna::ShiftMix((v ^ a) * mul);
return b;
}
// Return an 8-byte hash for 33 to 64 bytes.
STATIC_INLINE uint64_t HashLen33to64(const char *s, size_t len) {
uint64_t mul0 = k2 - 30;
uint64_t mul1 = k2 - 30 + 2 * len;
uint64_t h0 = H32(s, 32, mul0);
uint64_t h1 = H32(s + len - 32, 32, mul1);
return ((h1 * mul1) + h0) * mul1;
}
// Return an 8-byte hash for 65 to 96 bytes.
STATIC_INLINE uint64_t HashLen65to96(const char *s, size_t len) {
uint64_t mul0 = k2 - 114;
uint64_t mul1 = k2 - 114 + 2 * len;
uint64_t h0 = H32(s, 32, mul0);
uint64_t h1 = H32(s + 32, 32, mul1);
uint64_t h2 = H32(s + len - 32, 32, mul1, h0, h1);
return (h2 * 9 + (h0 >> 17) + (h1 >> 21)) * mul1;
}
uint64_t Hash64(const char *s, size_t len) {
if (len <= 32) {
if (len <= 16) {
return farmhashna::HashLen0to16(s, len);
} else {
return farmhashna::HashLen17to32(s, len);
}
} else if (len <= 64) {
return HashLen33to64(s, len);
} else if (len <= 96) {
return HashLen65to96(s, len);
} else if (len <= 256) {
return farmhashna::Hash64(s, len);
} else {
return farmhashuo::Hash64(s, len);
}
}
uint64_t Hash64WithSeeds(const char *s, size_t len, uint64_t seed0, uint64_t seed1) {
return farmhashuo::Hash64WithSeeds(s, len, seed0, seed1);
}
uint64_t Hash64WithSeed(const char *s, size_t len, uint64_t seed) {
return farmhashuo::Hash64WithSeed(s, len, seed);
}
} // namespace farmhashxo
namespace farmhashte {
#if !can_use_sse41 || !x86_64
uint64_t Hash64(const char *s, size_t len) {
FARMHASH_DIE_IF_MISCONFIGURED;
return s == NULL ? 0 : len;
}
uint64_t Hash64WithSeed(const char *s, size_t len, uint64_t seed) {
FARMHASH_DIE_IF_MISCONFIGURED;
return seed + Hash64(s, len);
}
uint64_t Hash64WithSeeds(const char *s, size_t len,
uint64_t seed0, uint64_t seed1) {
FARMHASH_DIE_IF_MISCONFIGURED;
return seed0 + seed1 + Hash64(s, len);
}
#else
#undef Fetch
#define Fetch Fetch64
#undef Rotate
#define Rotate Rotate64
#undef Bswap
#define Bswap Bswap64
// Helpers for data-parallel operations (1x 128 bits or 2x 64 or 4x 32).
STATIC_INLINE __m128i Add(__m128i x, __m128i y) { return _mm_add_epi64(x, y); }
STATIC_INLINE __m128i Xor(__m128i x, __m128i y) { return _mm_xor_si128(x, y); }
STATIC_INLINE __m128i Mul(__m128i x, __m128i y) { return _mm_mullo_epi32(x, y); }
STATIC_INLINE __m128i Shuf(__m128i x, __m128i y) { return _mm_shuffle_epi8(y, x); }
// Requires n >= 256. Requires SSE4.1. Should be slightly faster if the
// compiler uses AVX instructions (e.g., use the -mavx flag with GCC).
STATIC_INLINE uint64_t Hash64Long(const char* s, size_t n,
uint64_t seed0, uint64_t seed1) {
const __m128i kShuf =
_mm_set_epi8(4, 11, 10, 5, 8, 15, 6, 9, 12, 2, 14, 13, 0, 7, 3, 1);
const __m128i kMult =
_mm_set_epi8(0xbd, 0xd6, 0x33, 0x39, 0x45, 0x54, 0xfa, 0x03,
0x34, 0x3e, 0x33, 0xed, 0xcc, 0x9e, 0x2d, 0x51);
uint64_t seed2 = (seed0 + 113) * (seed1 + 9);
uint64_t seed3 = (Rotate(seed0, 23) + 27) * (Rotate(seed1, 30) + 111);
__m128i d0 = _mm_cvtsi64_si128(seed0);
__m128i d1 = _mm_cvtsi64_si128(seed1);
__m128i d2 = Shuf(kShuf, d0);
__m128i d3 = Shuf(kShuf, d1);
__m128i d4 = Xor(d0, d1);
__m128i d5 = Xor(d1, d2);
__m128i d6 = Xor(d2, d4);
__m128i d7 = _mm_set1_epi32(seed2 >> 32);
__m128i d8 = Mul(kMult, d2);
__m128i d9 = _mm_set1_epi32(seed3 >> 32);
__m128i d10 = _mm_set1_epi32(seed3);
__m128i d11 = Add(d2, _mm_set1_epi32(seed2));
const char* end = s + (n & ~static_cast<size_t>(255));
do {
__m128i z;
z = Fetch128(s);
d0 = Add(d0, z);
d1 = Shuf(kShuf, d1);
d2 = Xor(d2, d0);
d4 = Xor(d4, z);
d4 = Xor(d4, d1);
std::swap(d0, d6);
z = Fetch128(s + 16);
d5 = Add(d5, z);
d6 = Shuf(kShuf, d6);
d8 = Shuf(kShuf, d8);
d7 = Xor(d7, d5);
d0 = Xor(d0, z);
d0 = Xor(d0, d6);
std::swap(d5, d11);
z = Fetch128(s + 32);
d1 = Add(d1, z);
d2 = Shuf(kShuf, d2);
d4 = Shuf(kShuf, d4);
d5 = Xor(d5, z);
d5 = Xor(d5, d2);
std::swap(d10, d4);
z = Fetch128(s + 48);
d6 = Add(d6, z);
d7 = Shuf(kShuf, d7);
d0 = Shuf(kShuf, d0);
d8 = Xor(d8, d6);
d1 = Xor(d1, z);
d1 = Add(d1, d7);
z = Fetch128(s + 64);
d2 = Add(d2, z);
d5 = Shuf(kShuf, d5);
d4 = Add(d4, d2);
d6 = Xor(d6, z);
d6 = Xor(d6, d11);
std::swap(d8, d2);
z = Fetch128(s + 80);
d7 = Xor(d7, z);
d8 = Shuf(kShuf, d8);
d1 = Shuf(kShuf, d1);
d0 = Add(d0, d7);
d2 = Add(d2, z);
d2 = Add(d2, d8);
std::swap(d1, d7);
z = Fetch128(s + 96);
d4 = Shuf(kShuf, d4);
d6 = Shuf(kShuf, d6);
d8 = Mul(kMult, d8);
d5 = Xor(d5, d11);
d7 = Xor(d7, z);
d7 = Add(d7, d4);
std::swap(d6, d0);
z = Fetch128(s + 112);
d8 = Add(d8, z);
d0 = Shuf(kShuf, d0);
d2 = Shuf(kShuf, d2);
d1 = Xor(d1, d8);
d10 = Xor(d10, z);
d10 = Xor(d10, d0);
std::swap(d11, d5);
z = Fetch128(s + 128);
d4 = Add(d4, z);
d5 = Shuf(kShuf, d5);
d7 = Shuf(kShuf, d7);
d6 = Add(d6, d4);
d8 = Xor(d8, z);
d8 = Xor(d8, d5);
std::swap(d4, d10);
z = Fetch128(s + 144);
d0 = Add(d0, z);
d1 = Shuf(kShuf, d1);
d2 = Add(d2, d0);
d4 = Xor(d4, z);
d4 = Xor(d4, d1);
z = Fetch128(s + 160);
d5 = Add(d5, z);
d6 = Shuf(kShuf, d6);
d8 = Shuf(kShuf, d8);
d7 = Xor(d7, d5);
d0 = Xor(d0, z);
d0 = Xor(d0, d6);
std::swap(d2, d8);
z = Fetch128(s + 176);
d1 = Add(d1, z);
d2 = Shuf(kShuf, d2);
d4 = Shuf(kShuf, d4);
d5 = Mul(kMult, d5);
d5 = Xor(d5, z);
d5 = Xor(d5, d2);
std::swap(d7, d1);
z = Fetch128(s + 192);
d6 = Add(d6, z);
d7 = Shuf(kShuf, d7);
d0 = Shuf(kShuf, d0);
d8 = Add(d8, d6);
d1 = Xor(d1, z);
d1 = Xor(d1, d7);
std::swap(d0, d6);
z = Fetch128(s + 208);
d2 = Add(d2, z);
d5 = Shuf(kShuf, d5);
d4 = Xor(d4, d2);
d6 = Xor(d6, z);
d6 = Xor(d6, d9);
std::swap(d5, d11);
z = Fetch128(s + 224);
d7 = Add(d7, z);
d8 = Shuf(kShuf, d8);
d1 = Shuf(kShuf, d1);
d0 = Xor(d0, d7);
d2 = Xor(d2, z);
d2 = Xor(d2, d8);
std::swap(d10, d4);
z = Fetch128(s + 240);
d3 = Add(d3, z);
d4 = Shuf(kShuf, d4);
d6 = Shuf(kShuf, d6);
d7 = Mul(kMult, d7);
d5 = Add(d5, d3);
d7 = Xor(d7, z);
d7 = Xor(d7, d4);
std::swap(d3, d9);
s += 256;
} while (s != end);
d6 = Add(Mul(kMult, d6), _mm_cvtsi64_si128(n));
if (n % 256 != 0) {
d7 = Add(_mm_shuffle_epi32(d8, (0 << 6) + (3 << 4) + (2 << 2) + (1 << 0)), d7);
d8 = Add(Mul(kMult, d8), _mm_cvtsi64_si128(farmhashxo::Hash64(s, n % 256)));
}
__m128i t[8];
d0 = Mul(kMult, Shuf(kShuf, Mul(kMult, d0)));
d3 = Mul(kMult, Shuf(kShuf, Mul(kMult, d3)));
d9 = Mul(kMult, Shuf(kShuf, Mul(kMult, d9)));
d1 = Mul(kMult, Shuf(kShuf, Mul(kMult, d1)));
d0 = Add(d11, d0);
d3 = Xor(d7, d3);
d9 = Add(d8, d9);
d1 = Add(d10, d1);
d4 = Add(d3, d4);
d5 = Add(d9, d5);
d6 = Xor(d1, d6);
d2 = Add(d0, d2);
t[0] = d0;
t[1] = d3;
t[2] = d9;
t[3] = d1;
t[4] = d4;
t[5] = d5;
t[6] = d6;
t[7] = d2;
return farmhashxo::Hash64(reinterpret_cast<const char*>(t), sizeof(t));
}
uint64_t Hash64(const char *s, size_t len) {
// Empirically, farmhashxo seems faster until length 512.
return len >= 512 ? Hash64Long(s, len, k2, k1) : farmhashxo::Hash64(s, len);
}
uint64_t Hash64WithSeed(const char *s, size_t len, uint64_t seed) {
return len >= 512 ? Hash64Long(s, len, k1, seed) :
farmhashxo::Hash64WithSeed(s, len, seed);
}
uint64_t Hash64WithSeeds(const char *s, size_t len, uint64_t seed0, uint64_t seed1) {
return len >= 512 ? Hash64Long(s, len, seed0, seed1) :
farmhashxo::Hash64WithSeeds(s, len, seed0, seed1);
}