1 /* SPDX-License-Identifier: LicenseRef-lookup3-public-domain */
2 /* Slightly modified by Lennart Poettering, to avoid name clashes, and
3 * unexport a few functions. */
4
5 #include "lookup3.h"
6
7 /*
8 -------------------------------------------------------------------------------
9 lookup3.c, by Bob Jenkins, May 2006, Public Domain.
10
11 These are functions for producing 32-bit hashes for hash table lookup.
12 hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
13 are externally useful functions. Routines to test the hash are included
14 if SELF_TEST is defined. You can use this free for any purpose. It's in
15 the public domain. It has no warranty.
16
17 You probably want to use hashlittle(). hashlittle() and hashbig()
18 hash byte arrays. hashlittle() is faster than hashbig() on
19 little-endian machines. Intel and AMD are little-endian machines.
20 On second thought, you probably want hashlittle2(), which is identical to
21 hashlittle() except it returns two 32-bit hashes for the price of one.
22 You could implement hashbig2() if you wanted but I haven't bothered here.
23
24 If you want to find a hash of, say, exactly 7 integers, do
25 a = i1; b = i2; c = i3;
26 mix(a,b,c);
27 a += i4; b += i5; c += i6;
28 mix(a,b,c);
29 a += i7;
30 final(a,b,c);
31 then use c as the hash value. If you have a variable length array of
32 4-byte integers to hash, use hashword(). If you have a byte array (like
33 a character string), use hashlittle(). If you have several byte arrays, or
34 a mix of things, see the comments above hashlittle().
35
36 Why is this so big? I read 12 bytes at a time into 3 4-byte integers,
37 then mix those integers. This is fast (you can do a lot more thorough
38 mixing with 12*3 instructions on 3 integers than you can with 3 instructions
39 on 1 byte), but shoehorning those bytes into integers efficiently is messy.
40 -------------------------------------------------------------------------------
41 */
42 /* #define SELF_TEST 1 */
43
44 #include <stdint.h> /* defines uint32_t etc */
45 #include <stdio.h> /* defines printf for tests */
46 #include <sys/param.h> /* attempt to define endianness */
47 #include <time.h> /* defines time_t for timings in the test */
48 #ifdef linux
49 # include <endian.h> /* attempt to define endianness */
50 #endif
51
52 #if __GNUC__ >= 7
53 _Pragma("GCC diagnostic ignored \"-Wimplicit-fallthrough\"")
54 #endif
55
56 /*
57 * My best guess at if you are big-endian or little-endian. This may
58 * need adjustment.
59 */
60 #if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \
61 __BYTE_ORDER == __LITTLE_ENDIAN) || \
62 (defined(i386) || defined(__i386__) || defined(__i486__) || \
63 defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL))
64 # define HASH_LITTLE_ENDIAN 1
65 # define HASH_BIG_ENDIAN 0
66 #elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \
67 __BYTE_ORDER == __BIG_ENDIAN) || \
68 (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel))
69 # define HASH_LITTLE_ENDIAN 0
70 # define HASH_BIG_ENDIAN 1
71 #else
72 # define HASH_LITTLE_ENDIAN 0
73 # define HASH_BIG_ENDIAN 0
74 #endif
75
76 #define hashsize(n) ((uint32_t)1<<(n))
77 #define hashmask(n) (hashsize(n)-1)
78 #define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
79
80 /*
81 -------------------------------------------------------------------------------
82 mix -- mix 3 32-bit values reversibly.
83
84 This is reversible, so any information in (a,b,c) before mix() is
85 still in (a,b,c) after mix().
86
87 If four pairs of (a,b,c) inputs are run through mix(), or through
88 mix() in reverse, there are at least 32 bits of the output that
89 are sometimes the same for one pair and different for another pair.
90 This was tested for:
91 * pairs that differed by one bit, by two bits, in any combination
92 of top bits of (a,b,c), or in any combination of bottom bits of
93 (a,b,c).
94 * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
95 the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
96 is commonly produced by subtraction) look like a single 1-bit
97 difference.
98 * the base values were pseudorandom, all zero but one bit set, or
99 all zero plus a counter that starts at zero.
100
101 Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
102 satisfy this are
103 4 6 8 16 19 4
104 9 15 3 18 27 15
105 14 9 3 7 17 3
106 Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
107 for "differ" defined as + with a one-bit base and a two-bit delta. I
108 used http://burtleburtle.net/bob/hash/avalanche.html to choose
109 the operations, constants, and arrangements of the variables.
110
111 This does not achieve avalanche. There are input bits of (a,b,c)
112 that fail to affect some output bits of (a,b,c), especially of a. The
113 most thoroughly mixed value is c, but it doesn't really even achieve
114 avalanche in c.
115
116 This allows some parallelism. Read-after-writes are good at doubling
117 the number of bits affected, so the goal of mixing pulls in the opposite
118 direction as the goal of parallelism. I did what I could. Rotates
119 seem to cost as much as shifts on every machine I could lay my hands
120 on, and rotates are much kinder to the top and bottom bits, so I used
121 rotates.
122 -------------------------------------------------------------------------------
123 */
124 #define mix(a,b,c) \
125 { \
126 a -= c; a ^= rot(c, 4); c += b; \
127 b -= a; b ^= rot(a, 6); a += c; \
128 c -= b; c ^= rot(b, 8); b += a; \
129 a -= c; a ^= rot(c,16); c += b; \
130 b -= a; b ^= rot(a,19); a += c; \
131 c -= b; c ^= rot(b, 4); b += a; \
132 }
133
134 /*
135 -------------------------------------------------------------------------------
136 final -- final mixing of 3 32-bit values (a,b,c) into c
137
138 Pairs of (a,b,c) values differing in only a few bits will usually
139 produce values of c that look totally different. This was tested for
140 * pairs that differed by one bit, by two bits, in any combination
141 of top bits of (a,b,c), or in any combination of bottom bits of
142 (a,b,c).
143 * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
144 the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
145 is commonly produced by subtraction) look like a single 1-bit
146 difference.
147 * the base values were pseudorandom, all zero but one bit set, or
148 all zero plus a counter that starts at zero.
149
150 These constants passed:
151 14 11 25 16 4 14 24
152 12 14 25 16 4 14 24
153 and these came close:
154 4 8 15 26 3 22 24
155 10 8 15 26 3 22 24
156 11 8 15 26 3 22 24
157 -------------------------------------------------------------------------------
158 */
159 #define final(a,b,c) \
160 { \
161 c ^= b; c -= rot(b,14); \
162 a ^= c; a -= rot(c,11); \
163 b ^= a; b -= rot(a,25); \
164 c ^= b; c -= rot(b,16); \
165 a ^= c; a -= rot(c,4); \
166 b ^= a; b -= rot(a,14); \
167 c ^= b; c -= rot(b,24); \
168 }
169
170 /*
171 --------------------------------------------------------------------
172 This works on all machines. To be useful, it requires
173 -- that the key be an array of uint32_t's, and
174 -- that the length be the number of uint32_t's in the key
175
176 The function hashword() is identical to hashlittle() on little-endian
177 machines, and identical to hashbig() on big-endian machines,
178 except that the length has to be measured in uint32_ts rather than in
179 bytes. hashlittle() is more complicated than hashword() only because
180 hashlittle() has to dance around fitting the key bytes into registers.
181 --------------------------------------------------------------------
182 */
jenkins_hashword(const uint32_t * k,size_t length,uint32_t initval)183 uint32_t jenkins_hashword(
184 const uint32_t *k, /* the key, an array of uint32_t values */
185 size_t length, /* the length of the key, in uint32_ts */
186 uint32_t initval) /* the previous hash, or an arbitrary value */
187 {
188 uint32_t a,b,c;
189
190 /* Set up the internal state */
191 a = b = c = 0xdeadbeef + (((uint32_t)length)<<2) + initval;
192
193 /*------------------------------------------------- handle most of the key */
194 while (length > 3)
195 {
196 a += k[0];
197 b += k[1];
198 c += k[2];
199 mix(a,b,c);
200 length -= 3;
201 k += 3;
202 }
203
204 /*------------------------------------------- handle the last 3 uint32_t's */
205 switch(length) /* all the case statements fall through */
206 {
207 case 3 : c+=k[2];
208 case 2 : b+=k[1];
209 case 1 : a+=k[0];
210 final(a,b,c);
211 case 0: /* case 0: nothing left to add */
212 break;
213 }
214 /*------------------------------------------------------ report the result */
215 return c;
216 }
217
218 /*
219 --------------------------------------------------------------------
220 hashword2() -- same as hashword(), but take two seeds and return two
221 32-bit values. pc and pb must both be nonnull, and *pc and *pb must
222 both be initialized with seeds. If you pass in (*pb)==0, the output
223 (*pc) will be the same as the return value from hashword().
224 --------------------------------------------------------------------
225 */
jenkins_hashword2(const uint32_t * k,size_t length,uint32_t * pc,uint32_t * pb)226 void jenkins_hashword2 (
227 const uint32_t *k, /* the key, an array of uint32_t values */
228 size_t length, /* the length of the key, in uint32_ts */
229 uint32_t *pc, /* IN: seed OUT: primary hash value */
230 uint32_t *pb) /* IN: more seed OUT: secondary hash value */
231 {
232 uint32_t a,b,c;
233
234 /* Set up the internal state */
235 a = b = c = 0xdeadbeef + ((uint32_t)(length<<2)) + *pc;
236 c += *pb;
237
238 /*------------------------------------------------- handle most of the key */
239 while (length > 3)
240 {
241 a += k[0];
242 b += k[1];
243 c += k[2];
244 mix(a,b,c);
245 length -= 3;
246 k += 3;
247 }
248
249 /*------------------------------------------- handle the last 3 uint32_t's */
250 switch(length) /* all the case statements fall through */
251 {
252 case 3 : c+=k[2];
253 case 2 : b+=k[1];
254 case 1 : a+=k[0];
255 final(a,b,c);
256 case 0: /* case 0: nothing left to add */
257 break;
258 }
259 /*------------------------------------------------------ report the result */
260 *pc=c; *pb=b;
261 }
262
263 /*
264 -------------------------------------------------------------------------------
265 hashlittle() -- hash a variable-length key into a 32-bit value
266 k : the key (the unaligned variable-length array of bytes)
267 length : the length of the key, counting by bytes
268 initval : can be any 4-byte value
269 Returns a 32-bit value. Every bit of the key affects every bit of
270 the return value. Two keys differing by one or two bits will have
271 totally different hash values.
272
273 The best hash table sizes are powers of 2. There is no need to do
274 mod a prime (mod is sooo slow!). If you need less than 32 bits,
275 use a bitmask. For example, if you need only 10 bits, do
276 h = (h & hashmask(10));
277 In which case, the hash table should have hashsize(10) elements.
278
279 If you are hashing n strings (uint8_t **)k, do it like this:
280 for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
281
282 By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this
283 code any way you wish, private, educational, or commercial. It's free.
284
285 Use for hash table lookup, or anything where one collision in 2^^32 is
286 acceptable. Do NOT use for cryptographic purposes.
287 -------------------------------------------------------------------------------
288 */
289
jenkins_hashlittle(const void * key,size_t length,uint32_t initval)290 uint32_t jenkins_hashlittle( const void *key, size_t length, uint32_t initval)
291 {
292 uint32_t a,b,c; /* internal state */
293 union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */
294
295 /* Set up the internal state */
296 a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
297
298 u.ptr = key;
299 if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
300 const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
301
302 /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
303 while (length > 12)
304 {
305 a += k[0];
306 b += k[1];
307 c += k[2];
308 mix(a,b,c);
309 length -= 12;
310 k += 3;
311 }
312
313 /*----------------------------- handle the last (probably partial) block */
314 /*
315 * "k[2]&0xffffff" actually reads beyond the end of the string, but
316 * then masks off the part it's not allowed to read. Because the
317 * string is aligned, the masked-off tail is in the same word as the
318 * rest of the string. Every machine with memory protection I've seen
319 * does it on word boundaries, so is OK with this. But valgrind will
320 * still catch it and complain. The masking trick does make the hash
321 * noticeably faster for short strings (like English words).
322 */
323 #if !VALGRIND && !HAS_FEATURE_ADDRESS_SANITIZER && !HAS_FEATURE_MEMORY_SANITIZER
324
325 switch(length)
326 {
327 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
328 case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
329 case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
330 case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
331 case 8 : b+=k[1]; a+=k[0]; break;
332 case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
333 case 6 : b+=k[1]&0xffff; a+=k[0]; break;
334 case 5 : b+=k[1]&0xff; a+=k[0]; break;
335 case 4 : a+=k[0]; break;
336 case 3 : a+=k[0]&0xffffff; break;
337 case 2 : a+=k[0]&0xffff; break;
338 case 1 : a+=k[0]&0xff; break;
339 case 0 : return c; /* zero length strings require no mixing */
340 }
341
342 #else /* make valgrind happy */
343 {
344 const uint8_t *k8 = (const uint8_t *) k;
345
346 switch(length)
347 {
348 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
349 case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
350 case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
351 case 9 : c+=k8[8]; /* fall through */
352 case 8 : b+=k[1]; a+=k[0]; break;
353 case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
354 case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
355 case 5 : b+=k8[4]; /* fall through */
356 case 4 : a+=k[0]; break;
357 case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
358 case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
359 case 1 : a+=k8[0]; break;
360 case 0 : return c;
361 }
362 }
363
364 #endif /* !valgrind */
365
366 } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
367 const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */
368 const uint8_t *k8;
369
370 /*--------------- all but last block: aligned reads and different mixing */
371 while (length > 12)
372 {
373 a += k[0] + (((uint32_t)k[1])<<16);
374 b += k[2] + (((uint32_t)k[3])<<16);
375 c += k[4] + (((uint32_t)k[5])<<16);
376 mix(a,b,c);
377 length -= 12;
378 k += 6;
379 }
380
381 /*----------------------------- handle the last (probably partial) block */
382 k8 = (const uint8_t *)k;
383 switch(length)
384 {
385 case 12: c+=k[4]+(((uint32_t)k[5])<<16);
386 b+=k[2]+(((uint32_t)k[3])<<16);
387 a+=k[0]+(((uint32_t)k[1])<<16);
388 break;
389 case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
390 case 10: c+=k[4];
391 b+=k[2]+(((uint32_t)k[3])<<16);
392 a+=k[0]+(((uint32_t)k[1])<<16);
393 break;
394 case 9 : c+=k8[8]; /* fall through */
395 case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
396 a+=k[0]+(((uint32_t)k[1])<<16);
397 break;
398 case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
399 case 6 : b+=k[2];
400 a+=k[0]+(((uint32_t)k[1])<<16);
401 break;
402 case 5 : b+=k8[4]; /* fall through */
403 case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
404 break;
405 case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
406 case 2 : a+=k[0];
407 break;
408 case 1 : a+=k8[0];
409 break;
410 case 0 : return c; /* zero length requires no mixing */
411 }
412
413 } else { /* need to read the key one byte at a time */
414 const uint8_t *k = (const uint8_t *)key;
415
416 /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
417 while (length > 12)
418 {
419 a += k[0];
420 a += ((uint32_t)k[1])<<8;
421 a += ((uint32_t)k[2])<<16;
422 a += ((uint32_t)k[3])<<24;
423 b += k[4];
424 b += ((uint32_t)k[5])<<8;
425 b += ((uint32_t)k[6])<<16;
426 b += ((uint32_t)k[7])<<24;
427 c += k[8];
428 c += ((uint32_t)k[9])<<8;
429 c += ((uint32_t)k[10])<<16;
430 c += ((uint32_t)k[11])<<24;
431 mix(a,b,c);
432 length -= 12;
433 k += 12;
434 }
435
436 /*-------------------------------- last block: affect all 32 bits of (c) */
437 switch(length) /* all the case statements fall through */
438 {
439 case 12: c+=((uint32_t)k[11])<<24;
440 case 11: c+=((uint32_t)k[10])<<16;
441 case 10: c+=((uint32_t)k[9])<<8;
442 case 9 : c+=k[8];
443 case 8 : b+=((uint32_t)k[7])<<24;
444 case 7 : b+=((uint32_t)k[6])<<16;
445 case 6 : b+=((uint32_t)k[5])<<8;
446 case 5 : b+=k[4];
447 case 4 : a+=((uint32_t)k[3])<<24;
448 case 3 : a+=((uint32_t)k[2])<<16;
449 case 2 : a+=((uint32_t)k[1])<<8;
450 case 1 : a+=k[0];
451 break;
452 case 0 : return c;
453 }
454 }
455
456 final(a,b,c);
457 return c;
458 }
459
460 /*
461 * hashlittle2: return 2 32-bit hash values
462 *
463 * This is identical to hashlittle(), except it returns two 32-bit hash
464 * values instead of just one. This is good enough for hash table
465 * lookup with 2^^64 buckets, or if you want a second hash if you're not
466 * happy with the first, or if you want a probably-unique 64-bit ID for
467 * the key. *pc is better mixed than *pb, so use *pc first. If you want
468 * a 64-bit value do something like "*pc + (((uint64_t)*pb)<<32)".
469 */
jenkins_hashlittle2(const void * key,size_t length,uint32_t * pc,uint32_t * pb)470 void jenkins_hashlittle2(
471 const void *key, /* the key to hash */
472 size_t length, /* length of the key */
473 uint32_t *pc, /* IN: primary initval, OUT: primary hash */
474 uint32_t *pb) /* IN: secondary initval, OUT: secondary hash */
475 {
476 uint32_t a,b,c; /* internal state */
477 union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */
478
479 /* Set up the internal state */
480 a = b = c = 0xdeadbeef + ((uint32_t)length) + *pc;
481 c += *pb;
482
483 u.ptr = key;
484 if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
485 const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
486
487 /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
488 while (length > 12)
489 {
490 a += k[0];
491 b += k[1];
492 c += k[2];
493 mix(a,b,c);
494 length -= 12;
495 k += 3;
496 }
497
498 /*----------------------------- handle the last (probably partial) block */
499 /*
500 * "k[2]&0xffffff" actually reads beyond the end of the string, but
501 * then masks off the part it's not allowed to read. Because the
502 * string is aligned, the masked-off tail is in the same word as the
503 * rest of the string. Every machine with memory protection I've seen
504 * does it on word boundaries, so is OK with this. But valgrind will
505 * still catch it and complain. The masking trick does make the hash
506 * noticeably faster for short strings (like English words).
507 */
508 #if !VALGRIND && !HAS_FEATURE_ADDRESS_SANITIZER && !HAS_FEATURE_MEMORY_SANITIZER
509
510 switch(length)
511 {
512 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
513 case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
514 case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
515 case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
516 case 8 : b+=k[1]; a+=k[0]; break;
517 case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
518 case 6 : b+=k[1]&0xffff; a+=k[0]; break;
519 case 5 : b+=k[1]&0xff; a+=k[0]; break;
520 case 4 : a+=k[0]; break;
521 case 3 : a+=k[0]&0xffffff; break;
522 case 2 : a+=k[0]&0xffff; break;
523 case 1 : a+=k[0]&0xff; break;
524 case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */
525 }
526
527 #else /* make valgrind happy */
528
529 {
530 const uint8_t *k8 = (const uint8_t *)k;
531 switch(length)
532 {
533 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
534 case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
535 case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
536 case 9 : c+=k8[8]; /* fall through */
537 case 8 : b+=k[1]; a+=k[0]; break;
538 case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
539 case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
540 case 5 : b+=k8[4]; /* fall through */
541 case 4 : a+=k[0]; break;
542 case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
543 case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
544 case 1 : a+=k8[0]; break;
545 case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */
546 }
547 }
548
549 #endif /* !valgrind */
550
551 } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
552 const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */
553 const uint8_t *k8;
554
555 /*--------------- all but last block: aligned reads and different mixing */
556 while (length > 12)
557 {
558 a += k[0] + (((uint32_t)k[1])<<16);
559 b += k[2] + (((uint32_t)k[3])<<16);
560 c += k[4] + (((uint32_t)k[5])<<16);
561 mix(a,b,c);
562 length -= 12;
563 k += 6;
564 }
565
566 /*----------------------------- handle the last (probably partial) block */
567 k8 = (const uint8_t *)k;
568 switch(length)
569 {
570 case 12: c+=k[4]+(((uint32_t)k[5])<<16);
571 b+=k[2]+(((uint32_t)k[3])<<16);
572 a+=k[0]+(((uint32_t)k[1])<<16);
573 break;
574 case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
575 case 10: c+=k[4];
576 b+=k[2]+(((uint32_t)k[3])<<16);
577 a+=k[0]+(((uint32_t)k[1])<<16);
578 break;
579 case 9 : c+=k8[8]; /* fall through */
580 case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
581 a+=k[0]+(((uint32_t)k[1])<<16);
582 break;
583 case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
584 case 6 : b+=k[2];
585 a+=k[0]+(((uint32_t)k[1])<<16);
586 break;
587 case 5 : b+=k8[4]; /* fall through */
588 case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
589 break;
590 case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
591 case 2 : a+=k[0];
592 break;
593 case 1 : a+=k8[0];
594 break;
595 case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */
596 }
597
598 } else { /* need to read the key one byte at a time */
599 const uint8_t *k = (const uint8_t *)key;
600
601 /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
602 while (length > 12)
603 {
604 a += k[0];
605 a += ((uint32_t)k[1])<<8;
606 a += ((uint32_t)k[2])<<16;
607 a += ((uint32_t)k[3])<<24;
608 b += k[4];
609 b += ((uint32_t)k[5])<<8;
610 b += ((uint32_t)k[6])<<16;
611 b += ((uint32_t)k[7])<<24;
612 c += k[8];
613 c += ((uint32_t)k[9])<<8;
614 c += ((uint32_t)k[10])<<16;
615 c += ((uint32_t)k[11])<<24;
616 mix(a,b,c);
617 length -= 12;
618 k += 12;
619 }
620
621 /*-------------------------------- last block: affect all 32 bits of (c) */
622 switch(length) /* all the case statements fall through */
623 {
624 case 12: c+=((uint32_t)k[11])<<24;
625 case 11: c+=((uint32_t)k[10])<<16;
626 case 10: c+=((uint32_t)k[9])<<8;
627 case 9 : c+=k[8];
628 case 8 : b+=((uint32_t)k[7])<<24;
629 case 7 : b+=((uint32_t)k[6])<<16;
630 case 6 : b+=((uint32_t)k[5])<<8;
631 case 5 : b+=k[4];
632 case 4 : a+=((uint32_t)k[3])<<24;
633 case 3 : a+=((uint32_t)k[2])<<16;
634 case 2 : a+=((uint32_t)k[1])<<8;
635 case 1 : a+=k[0];
636 break;
637 case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */
638 }
639 }
640
641 final(a,b,c);
642 *pc=c; *pb=b;
643 }
644
645 /*
646 * hashbig():
647 * This is the same as hashword() on big-endian machines. It is different
648 * from hashlittle() on all machines. hashbig() takes advantage of
649 * big-endian byte ordering.
650 */
jenkins_hashbig(const void * key,size_t length,uint32_t initval)651 uint32_t jenkins_hashbig( const void *key, size_t length, uint32_t initval)
652 {
653 uint32_t a,b,c;
654 union { const void *ptr; size_t i; } u; /* to cast key to (size_t) happily */
655
656 /* Set up the internal state */
657 a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
658
659 u.ptr = key;
660 if (HASH_BIG_ENDIAN && ((u.i & 0x3) == 0)) {
661 const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
662
663 /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
664 while (length > 12)
665 {
666 a += k[0];
667 b += k[1];
668 c += k[2];
669 mix(a,b,c);
670 length -= 12;
671 k += 3;
672 }
673
674 /*----------------------------- handle the last (probably partial) block */
675 /*
676 * "k[2]<<8" actually reads beyond the end of the string, but
677 * then shifts out the part it's not allowed to read. Because the
678 * string is aligned, the illegal read is in the same word as the
679 * rest of the string. Every machine with memory protection I've seen
680 * does it on word boundaries, so is OK with this. But valgrind will
681 * still catch it and complain. The masking trick does make the hash
682 * noticeably faster for short strings (like English words).
683 */
684 #if !VALGRIND && !HAS_FEATURE_ADDRESS_SANITIZER && !HAS_FEATURE_MEMORY_SANITIZER
685
686 switch(length)
687 {
688 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
689 case 11: c+=k[2]&0xffffff00; b+=k[1]; a+=k[0]; break;
690 case 10: c+=k[2]&0xffff0000; b+=k[1]; a+=k[0]; break;
691 case 9 : c+=k[2]&0xff000000; b+=k[1]; a+=k[0]; break;
692 case 8 : b+=k[1]; a+=k[0]; break;
693 case 7 : b+=k[1]&0xffffff00; a+=k[0]; break;
694 case 6 : b+=k[1]&0xffff0000; a+=k[0]; break;
695 case 5 : b+=k[1]&0xff000000; a+=k[0]; break;
696 case 4 : a+=k[0]; break;
697 case 3 : a+=k[0]&0xffffff00; break;
698 case 2 : a+=k[0]&0xffff0000; break;
699 case 1 : a+=k[0]&0xff000000; break;
700 case 0 : return c; /* zero length strings require no mixing */
701 }
702
703 #else /* make valgrind happy */
704
705 {
706 const uint8_t *k8 = (const uint8_t *)k;
707 switch(length) /* all the case statements fall through */
708 {
709 case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
710 case 11: c+=((uint32_t)k8[10])<<8; /* fall through */
711 case 10: c+=((uint32_t)k8[9])<<16; /* fall through */
712 case 9 : c+=((uint32_t)k8[8])<<24; /* fall through */
713 case 8 : b+=k[1]; a+=k[0]; break;
714 case 7 : b+=((uint32_t)k8[6])<<8; /* fall through */
715 case 6 : b+=((uint32_t)k8[5])<<16; /* fall through */
716 case 5 : b+=((uint32_t)k8[4])<<24; /* fall through */
717 case 4 : a+=k[0]; break;
718 case 3 : a+=((uint32_t)k8[2])<<8; /* fall through */
719 case 2 : a+=((uint32_t)k8[1])<<16; /* fall through */
720 case 1 : a+=((uint32_t)k8[0])<<24; break;
721 case 0 : return c;
722 }
723 }
724
725 #endif /* !VALGRIND */
726
727 } else { /* need to read the key one byte at a time */
728 const uint8_t *k = (const uint8_t *)key;
729
730 /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
731 while (length > 12)
732 {
733 a += ((uint32_t)k[0])<<24;
734 a += ((uint32_t)k[1])<<16;
735 a += ((uint32_t)k[2])<<8;
736 a += ((uint32_t)k[3]);
737 b += ((uint32_t)k[4])<<24;
738 b += ((uint32_t)k[5])<<16;
739 b += ((uint32_t)k[6])<<8;
740 b += ((uint32_t)k[7]);
741 c += ((uint32_t)k[8])<<24;
742 c += ((uint32_t)k[9])<<16;
743 c += ((uint32_t)k[10])<<8;
744 c += ((uint32_t)k[11]);
745 mix(a,b,c);
746 length -= 12;
747 k += 12;
748 }
749
750 /*-------------------------------- last block: affect all 32 bits of (c) */
751 switch(length) /* all the case statements fall through */
752 {
753 case 12: c+=k[11];
754 case 11: c+=((uint32_t)k[10])<<8;
755 case 10: c+=((uint32_t)k[9])<<16;
756 case 9 : c+=((uint32_t)k[8])<<24;
757 case 8 : b+=k[7];
758 case 7 : b+=((uint32_t)k[6])<<8;
759 case 6 : b+=((uint32_t)k[5])<<16;
760 case 5 : b+=((uint32_t)k[4])<<24;
761 case 4 : a+=k[3];
762 case 3 : a+=((uint32_t)k[2])<<8;
763 case 2 : a+=((uint32_t)k[1])<<16;
764 case 1 : a+=((uint32_t)k[0])<<24;
765 break;
766 case 0 : return c;
767 }
768 }
769
770 final(a,b,c);
771 return c;
772 }
773
774 #ifdef SELF_TEST
775
776 /* used for timings */
driver1()777 void driver1()
778 {
779 uint8_t buf[256];
780 uint32_t i;
781 uint32_t h=0;
782 time_t a,z;
783
784 time(&a);
785 for (i=0; i<256; ++i) buf[i] = 'x';
786 for (i=0; i<1; ++i)
787 {
788 h = hashlittle(&buf[0],1,h);
789 }
790 time(&z);
791 if (z-a > 0) printf("time %d %.8x\n", z-a, h);
792 }
793
794 /* check that every input bit changes every output bit half the time */
795 #define HASHSTATE 1
796 #define HASHLEN 1
797 #define MAXPAIR 60
798 #define MAXLEN 70
driver2()799 void driver2()
800 {
801 uint8_t qa[MAXLEN+1], qb[MAXLEN+2], *a = &qa[0], *b = &qb[1];
802 uint32_t c[HASHSTATE], d[HASHSTATE], i=0, j=0, k, l, m=0, z;
803 uint32_t e[HASHSTATE],f[HASHSTATE],g[HASHSTATE],h[HASHSTATE];
804 uint32_t x[HASHSTATE],y[HASHSTATE];
805 uint32_t hlen;
806
807 printf("No more than %d trials should ever be needed \n",MAXPAIR/2);
808 for (hlen=0; hlen < MAXLEN; ++hlen)
809 {
810 z=0;
811 for (i=0; i<hlen; ++i) /*----------------------- for each input byte, */
812 {
813 for (j=0; j<8; ++j) /*------------------------ for each input bit, */
814 {
815 for (m=1; m<8; ++m) /*------------- for several possible initvals, */
816 {
817 for (l=0; l<HASHSTATE; ++l)
818 e[l]=f[l]=g[l]=h[l]=x[l]=y[l]=~((uint32_t)0);
819
820 /*---- check that every output bit is affected by that input bit */
821 for (k=0; k<MAXPAIR; k+=2)
822 {
823 uint32_t finished=1;
824 /* keys have one bit different */
825 for (l=0; l<hlen+1; ++l) {a[l] = b[l] = (uint8_t)0;}
826 /* have a and b be two keys differing in only one bit */
827 a[i] ^= (k<<j);
828 a[i] ^= (k>>(8-j));
829 c[0] = hashlittle(a, hlen, m);
830 b[i] ^= ((k+1)<<j);
831 b[i] ^= ((k+1)>>(8-j));
832 d[0] = hashlittle(b, hlen, m);
833 /* check every bit is 1, 0, set, and not set at least once */
834 for (l=0; l<HASHSTATE; ++l)
835 {
836 e[l] &= (c[l]^d[l]);
837 f[l] &= ~(c[l]^d[l]);
838 g[l] &= c[l];
839 h[l] &= ~c[l];
840 x[l] &= d[l];
841 y[l] &= ~d[l];
842 if (e[l]|f[l]|g[l]|h[l]|x[l]|y[l]) finished=0;
843 }
844 if (finished) break;
845 }
846 if (k>z) z=k;
847 if (k==MAXPAIR)
848 {
849 printf("Some bit didn't change: ");
850 printf("%.8x %.8x %.8x %.8x %.8x %.8x ",
851 e[0],f[0],g[0],h[0],x[0],y[0]);
852 printf("i %d j %d m %d len %d\n", i, j, m, hlen);
853 }
854 if (z==MAXPAIR) goto done;
855 }
856 }
857 }
858 done:
859 if (z < MAXPAIR)
860 {
861 printf("Mix success %2d bytes %2d initvals ",i,m);
862 printf("required %d trials\n", z/2);
863 }
864 }
865 printf("\n");
866 }
867
868 /* Check for reading beyond the end of the buffer and alignment problems */
driver3()869 void driver3()
870 {
871 uint8_t buf[MAXLEN+20], *b;
872 uint32_t len;
873 uint8_t q[] = "This is the time for all good men to come to the aid of their country...";
874 uint32_t h;
875 uint8_t qq[] = "xThis is the time for all good men to come to the aid of their country...";
876 uint32_t i;
877 uint8_t qqq[] = "xxThis is the time for all good men to come to the aid of their country...";
878 uint32_t j;
879 uint8_t qqqq[] = "xxxThis is the time for all good men to come to the aid of their country...";
880 uint32_t ref,x,y;
881 uint8_t *p;
882
883 printf("Endianness. These lines should all be the same (for values filled in):\n");
884 printf("%.8x %.8x %.8x\n",
885 hashword((const uint32_t *)q, (sizeof(q)-1)/4, 13),
886 hashword((const uint32_t *)q, (sizeof(q)-5)/4, 13),
887 hashword((const uint32_t *)q, (sizeof(q)-9)/4, 13));
888 p = q;
889 printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
890 hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
891 hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
892 hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
893 hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
894 hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
895 hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
896 p = &qq[1];
897 printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
898 hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
899 hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
900 hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
901 hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
902 hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
903 hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
904 p = &qqq[2];
905 printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
906 hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
907 hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
908 hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
909 hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
910 hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
911 hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
912 p = &qqqq[3];
913 printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
914 hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
915 hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
916 hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
917 hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
918 hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
919 hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
920 printf("\n");
921
922 /* check that hashlittle2 and hashlittle produce the same results */
923 i=47; j=0;
924 hashlittle2(q, sizeof(q), &i, &j);
925 if (hashlittle(q, sizeof(q), 47) != i)
926 printf("hashlittle2 and hashlittle mismatch\n");
927
928 /* check that hashword2 and hashword produce the same results */
929 len = 0xdeadbeef;
930 i=47, j=0;
931 hashword2(&len, 1, &i, &j);
932 if (hashword(&len, 1, 47) != i)
933 printf("hashword2 and hashword mismatch %x %x\n",
934 i, hashword(&len, 1, 47));
935
936 /* check hashlittle doesn't read before or after the ends of the string */
937 for (h=0, b=buf+1; h<8; ++h, ++b)
938 {
939 for (i=0; i<MAXLEN; ++i)
940 {
941 len = i;
942 for (j=0; j<i; ++j) *(b+j)=0;
943
944 /* these should all be equal */
945 ref = hashlittle(b, len, (uint32_t)1);
946 *(b+i)=(uint8_t)~0;
947 *(b-1)=(uint8_t)~0;
948 x = hashlittle(b, len, (uint32_t)1);
949 y = hashlittle(b, len, (uint32_t)1);
950 if ((ref != x) || (ref != y))
951 {
952 printf("alignment error: %.8x %.8x %.8x %d %d\n",ref,x,y,
953 h, i);
954 }
955 }
956 }
957 }
958
959 /* check for problems with nulls */
driver4()960 void driver4()
961 {
962 uint8_t buf[1];
963 uint32_t h,i,state[HASHSTATE];
964
965 buf[0] = ~0;
966 for (i=0; i<HASHSTATE; ++i) state[i] = 1;
967 printf("These should all be different\n");
968 for (i=0, h=0; i<8; ++i)
969 {
970 h = hashlittle(buf, 0, h);
971 printf("%2ld 0-byte strings, hash is %.8x\n", i, h);
972 }
973 }
974
driver5()975 void driver5()
976 {
977 uint32_t b,c;
978 b=0, c=0, hashlittle2("", 0, &c, &b);
979 printf("hash is %.8lx %.8lx\n", c, b); /* deadbeef deadbeef */
980 b=0xdeadbeef, c=0, hashlittle2("", 0, &c, &b);
981 printf("hash is %.8lx %.8lx\n", c, b); /* bd5b7dde deadbeef */
982 b=0xdeadbeef, c=0xdeadbeef, hashlittle2("", 0, &c, &b);
983 printf("hash is %.8lx %.8lx\n", c, b); /* 9c093ccd bd5b7dde */
984 b=0, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b);
985 printf("hash is %.8lx %.8lx\n", c, b); /* 17770551 ce7226e6 */
986 b=1, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b);
987 printf("hash is %.8lx %.8lx\n", c, b); /* e3607cae bd371de4 */
988 b=0, c=1, hashlittle2("Four score and seven years ago", 30, &c, &b);
989 printf("hash is %.8lx %.8lx\n", c, b); /* cd628161 6cbea4b3 */
990 c = hashlittle("Four score and seven years ago", 30, 0);
991 printf("hash is %.8lx\n", c); /* 17770551 */
992 c = hashlittle("Four score and seven years ago", 30, 1);
993 printf("hash is %.8lx\n", c); /* cd628161 */
994 }
995
main()996 int main()
997 {
998 driver1(); /* test that the key is hashed: used for timings */
999 driver2(); /* test that whole key is hashed thoroughly */
1000 driver3(); /* test that nothing but the key is hashed */
1001 driver4(); /* test hashing multiple buffers (all buffers are null) */
1002 driver5(); /* test the hash against known vectors */
1003 return 1;
1004 }
1005
1006 #endif /* SELF_TEST */
1007