1/*
2-------------------------------------------------------------------------------
3lookup3.c, by Bob Jenkins, May 2006, Public Domain.
4
5These are functions for producing 32-bit hashes for hash table lookup.
6hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final()
7are externally useful functions.  Routines to test the hash are included
8if SELF_TEST is defined.  You can use this free for any purpose.  It's in
9the public domain.  It has no warranty.
10
11You probably want to use hashlittle().  hashlittle() and hashbig()
12hash byte arrays.  hashlittle() is is faster than hashbig() on
13little-endian machines.  Intel and AMD are little-endian machines.
14On second thought, you probably want hashlittle2(), which is identical to
15hashlittle() except it returns two 32-bit hashes for the price of one.
16You could implement hashbig2() if you wanted but I haven't bothered here.
17
18If you want to find a hash of, say, exactly 7 integers, do
19  a = i1;  b = i2;  c = i3;
20  mix(a,b,c);
21  a += i4; b += i5; c += i6;
22  mix(a,b,c);
23  a += i7;
24  final(a,b,c);
25then use c as the hash value.  If you have a variable length array of
264-byte integers to hash, use hashword().  If you have a byte array (like
27a character string), use hashlittle().  If you have several byte arrays, or
28a mix of things, see the comments above hashlittle().
29
30Why is this so big?  I read 12 bytes at a time into 3 4-byte integers,
31then mix those integers.  This is fast (you can do a lot more thorough
32mixing with 12*3 instructions on 3 integers than you can with 3 instructions
33on 1 byte), but shoehorning those bytes into integers efficiently is messy.
34-------------------------------------------------------------------------------
35*/
36#define SELF_TEST 1
37#undef SELF_TEST
38
39#include <stdio.h>      /* defines printf for tests */
40#include <time.h>       /* defines time_t for timings in the test */
41#include <stdint.h>     /* defines uint32_t etc */
42#include <sys/param.h>  /* attempt to define endianness */
43#ifdef linux
44# include <endian.h>    /* attempt to define endianness */
45#endif
46
47/*
48 * My best guess at if you are big-endian or little-endian.  This may
49 * need adjustment.
50 */
51#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \
52     __BYTE_ORDER == __LITTLE_ENDIAN) || \
53    (defined(i386) || defined(__i386__) || defined(__i486__) || \
54     defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL))
55# define HASH_LITTLE_ENDIAN 1
56# define HASH_BIG_ENDIAN 0
57#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \
58       __BYTE_ORDER == __BIG_ENDIAN) || \
59      (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel))
60# define HASH_LITTLE_ENDIAN 0
61# define HASH_BIG_ENDIAN 1
62#else
63# define HASH_LITTLE_ENDIAN 0
64# define HASH_BIG_ENDIAN 0
65#endif
66
67#define hashsize(n) ((uint32_t)1<<(n))
68#define hashmask(n) (hashsize(n)-1)
69#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
70
71/*
72-------------------------------------------------------------------------------
73mix -- mix 3 32-bit values reversibly.
74
75This is reversible, so any information in (a,b,c) before mix() is
76still in (a,b,c) after mix().
77
78If four pairs of (a,b,c) inputs are run through mix(), or through
79mix() in reverse, there are at least 32 bits of the output that
80are sometimes the same for one pair and different for another pair.
81This was tested for:
82* pairs that differed by one bit, by two bits, in any combination
83  of top bits of (a,b,c), or in any combination of bottom bits of
84  (a,b,c).
85* "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
86  the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
87  is commonly produced by subtraction) look like a single 1-bit
88  difference.
89* the base values were pseudorandom, all zero but one bit set, or
90  all zero plus a counter that starts at zero.
91
92Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
93satisfy this are
94    4  6  8 16 19  4
95    9 15  3 18 27 15
96   14  9  3  7 17  3
97Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
98for "differ" defined as + with a one-bit base and a two-bit delta.  I
99used http://burtleburtle.net/bob/hash/avalanche.html to choose
100the operations, constants, and arrangements of the variables.
101
102This does not achieve avalanche.  There are input bits of (a,b,c)
103that fail to affect some output bits of (a,b,c), especially of a.  The
104most thoroughly mixed value is c, but it doesn't really even achieve
105avalanche in c.
106
107This allows some parallelism.  Read-after-writes are good at doubling
108the number of bits affected, so the goal of mixing pulls in the opposite
109direction as the goal of parallelism.  I did what I could.  Rotates
110seem to cost as much as shifts on every machine I could lay my hands
111on, and rotates are much kinder to the top and bottom bits, so I used
112rotates.
113-------------------------------------------------------------------------------
114*/
115#define mix(a,b,c) \
116{ \
117  a -= c;  a ^= rot(c, 4);  c += b; \
118  b -= a;  b ^= rot(a, 6);  a += c; \
119  c -= b;  c ^= rot(b, 8);  b += a; \
120  a -= c;  a ^= rot(c,16);  c += b; \
121  b -= a;  b ^= rot(a,19);  a += c; \
122  c -= b;  c ^= rot(b, 4);  b += a; \
123}
124
125/*
126-------------------------------------------------------------------------------
127final -- final mixing of 3 32-bit values (a,b,c) into c
128
129Pairs of (a,b,c) values differing in only a few bits will usually
130produce values of c that look totally different.  This was tested for
131* pairs that differed by one bit, by two bits, in any combination
132  of top bits of (a,b,c), or in any combination of bottom bits of
133  (a,b,c).
134* "differ" is defined as +, -, ^, or ~^.  For + and -, I transformed
135  the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
136  is commonly produced by subtraction) look like a single 1-bit
137  difference.
138* the base values were pseudorandom, all zero but one bit set, or
139  all zero plus a counter that starts at zero.
140
141These constants passed:
142 14 11 25 16 4 14 24
143 12 14 25 16 4 14 24
144and these came close:
145  4  8 15 26 3 22 24
146 10  8 15 26 3 22 24
147 11  8 15 26 3 22 24
148-------------------------------------------------------------------------------
149*/
150#define final(a,b,c) \
151{ \
152  c ^= b; c -= rot(b,14); \
153  a ^= c; a -= rot(c,11); \
154  b ^= a; b -= rot(a,25); \
155  c ^= b; c -= rot(b,16); \
156  a ^= c; a -= rot(c,4);  \
157  b ^= a; b -= rot(a,14); \
158  c ^= b; c -= rot(b,24); \
159}
160
161/*
162--------------------------------------------------------------------
163 This works on all machines.  To be useful, it requires
164 -- that the key be an array of uint32_t's, and
165 -- that the length be the number of uint32_t's in the key
166
167 The function hashword() is identical to hashlittle() on little-endian
168 machines, and identical to hashbig() on big-endian machines,
169 except that the length has to be measured in uint32_ts rather than in
170 bytes.  hashlittle() is more complicated than hashword() only because
171 hashlittle() has to dance around fitting the key bytes into registers.
172--------------------------------------------------------------------
173*/
174uint32_t hashword(
175const uint32_t *k,                   /* the key, an array of uint32_t values */
176size_t          length,               /* the length of the key, in uint32_ts */
177uint32_t        initval)         /* the previous hash, or an arbitrary value */
178{
179  uint32_t a,b,c;
180
181  /* Set up the internal state */
182  a = b = c = 0xdeadbeef + (((uint32_t)length)<<2) + initval;
183
184  /*------------------------------------------------- handle most of the key */
185  while (length > 3)
186  {
187    a += k[0];
188    b += k[1];
189    c += k[2];
190    mix(a,b,c);
191    length -= 3;
192    k += 3;
193  }
194
195  /*------------------------------------------- handle the last 3 uint32_t's */
196  switch(length)                     /* all the case statements fall through */
197  {
198  case 3 : c+=k[2];
199  case 2 : b+=k[1];
200  case 1 : a+=k[0];
201    final(a,b,c);
202  case 0:     /* case 0: nothing left to add */
203    break;
204  }
205  /*------------------------------------------------------ report the result */
206  return c;
207}
208
209
210/*
211--------------------------------------------------------------------
212hashword2() -- same as hashword(), but take two seeds and return two
21332-bit values.  pc and pb must both be nonnull, and *pc and *pb must
214both be initialized with seeds.  If you pass in (*pb)==0, the output
215(*pc) will be the same as the return value from hashword().
216--------------------------------------------------------------------
217*/
218void hashword2 (
219const uint32_t *k,                   /* the key, an array of uint32_t values */
220size_t          length,               /* the length of the key, in uint32_ts */
221uint32_t       *pc,                      /* IN: seed OUT: primary hash value */
222uint32_t       *pb)               /* IN: more seed OUT: secondary hash value */
223{
224  uint32_t a,b,c;
225
226  /* Set up the internal state */
227  a = b = c = 0xdeadbeef + ((uint32_t)(length<<2)) + *pc;
228  c += *pb;
229
230  /*------------------------------------------------- handle most of the key */
231  while (length > 3)
232  {
233    a += k[0];
234    b += k[1];
235    c += k[2];
236    mix(a,b,c);
237    length -= 3;
238    k += 3;
239  }
240
241  /*------------------------------------------- handle the last 3 uint32_t's */
242  switch(length)                     /* all the case statements fall through */
243  {
244  case 3 : c+=k[2];
245  case 2 : b+=k[1];
246  case 1 : a+=k[0];
247    final(a,b,c);
248  case 0:     /* case 0: nothing left to add */
249    break;
250  }
251  /*------------------------------------------------------ report the result */
252  *pc=c; *pb=b;
253}
254
255
256/*
257-------------------------------------------------------------------------------
258hashlittle() -- hash a variable-length key into a 32-bit value
259  k       : the key (the unaligned variable-length array of bytes)
260  length  : the length of the key, counting by bytes
261  initval : can be any 4-byte value
262Returns a 32-bit value.  Every bit of the key affects every bit of
263the return value.  Two keys differing by one or two bits will have
264totally different hash values.
265
266The best hash table sizes are powers of 2.  There is no need to do
267mod a prime (mod is sooo slow!).  If you need less than 32 bits,
268use a bitmask.  For example, if you need only 10 bits, do
269  h = (h & hashmask(10));
270In which case, the hash table should have hashsize(10) elements.
271
272If you are hashing n strings (uint8_t **)k, do it like this:
273  for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
274
275By Bob Jenkins, 2006.  bob_jenkins@burtleburtle.net.  You may use this
276code any way you wish, private, educational, or commercial.  It's free.
277
278Use for hash table lookup, or anything where one collision in 2^^32 is
279acceptable.  Do NOT use for cryptographic purposes.
280-------------------------------------------------------------------------------
281*/
282
283uint32_t hashlittle( const void *key, size_t length, uint32_t initval)
284{
285  uint32_t a,b,c;                                          /* internal state */
286  union { const void *ptr; size_t i; } u;     /* needed for Mac Powerbook G4 */
287
288  /* Set up the internal state */
289  a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
290
291  u.ptr = key;
292  if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
293    const uint32_t *k = (const uint32_t *)key;         /* read 32-bit chunks */
294    const uint8_t  *k8;
295
296    /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
297    while (length > 12)
298    {
299      a += k[0];
300      b += k[1];
301      c += k[2];
302      mix(a,b,c);
303      length -= 12;
304      k += 3;
305    }
306
307    /*----------------------------- handle the last (probably partial) block */
308    /*
309     * "k[2]&0xffffff" actually reads beyond the end of the string, but
310     * then masks off the part it's not allowed to read.  Because the
311     * string is aligned, the masked-off tail is in the same word as the
312     * rest of the string.  Every machine with memory protection I've seen
313     * does it on word boundaries, so is OK with this.  But VALGRIND will
314     * still catch it and complain.  The masking trick does make the hash
315     * noticably faster for short strings (like English words).
316     */
317#ifndef VALGRIND
318
319    switch(length)
320    {
321    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
322    case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
323    case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
324    case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
325    case 8 : b+=k[1]; a+=k[0]; break;
326    case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
327    case 6 : b+=k[1]&0xffff; a+=k[0]; break;
328    case 5 : b+=k[1]&0xff; a+=k[0]; break;
329    case 4 : a+=k[0]; break;
330    case 3 : a+=k[0]&0xffffff; break;
331    case 2 : a+=k[0]&0xffff; break;
332    case 1 : a+=k[0]&0xff; break;
333    case 0 : return c;              /* zero length strings require no mixing */
334    }
335
336#else /* make valgrind happy */
337
338    k8 = (const uint8_t *)k;
339    switch(length)
340    {
341    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
342    case 11: c+=((uint32_t)k8[10])<<16;  /* fall through */
343    case 10: c+=((uint32_t)k8[9])<<8;    /* fall through */
344    case 9 : c+=k8[8];                   /* fall through */
345    case 8 : b+=k[1]; a+=k[0]; break;
346    case 7 : b+=((uint32_t)k8[6])<<16;   /* fall through */
347    case 6 : b+=((uint32_t)k8[5])<<8;    /* fall through */
348    case 5 : b+=k8[4];                   /* fall through */
349    case 4 : a+=k[0]; break;
350    case 3 : a+=((uint32_t)k8[2])<<16;   /* fall through */
351    case 2 : a+=((uint32_t)k8[1])<<8;    /* fall through */
352    case 1 : a+=k8[0]; break;
353    case 0 : return c;
354    }
355
356#endif /* !valgrind */
357
358  } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
359    const uint16_t *k = (const uint16_t *)key;         /* read 16-bit chunks */
360    const uint8_t  *k8;
361
362    /*--------------- all but last block: aligned reads and different mixing */
363    while (length > 12)
364    {
365      a += k[0] + (((uint32_t)k[1])<<16);
366      b += k[2] + (((uint32_t)k[3])<<16);
367      c += k[4] + (((uint32_t)k[5])<<16);
368      mix(a,b,c);
369      length -= 12;
370      k += 6;
371    }
372
373    /*----------------------------- handle the last (probably partial) block */
374    k8 = (const uint8_t *)k;
375    switch(length)
376    {
377    case 12: c+=k[4]+(((uint32_t)k[5])<<16);
378             b+=k[2]+(((uint32_t)k[3])<<16);
379             a+=k[0]+(((uint32_t)k[1])<<16);
380             break;
381    case 11: c+=((uint32_t)k8[10])<<16;     /* fall through */
382    case 10: c+=k[4];
383             b+=k[2]+(((uint32_t)k[3])<<16);
384             a+=k[0]+(((uint32_t)k[1])<<16);
385             break;
386    case 9 : c+=k8[8];                      /* fall through */
387    case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
388             a+=k[0]+(((uint32_t)k[1])<<16);
389             break;
390    case 7 : b+=((uint32_t)k8[6])<<16;      /* fall through */
391    case 6 : b+=k[2];
392             a+=k[0]+(((uint32_t)k[1])<<16);
393             break;
394    case 5 : b+=k8[4];                      /* fall through */
395    case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
396             break;
397    case 3 : a+=((uint32_t)k8[2])<<16;      /* fall through */
398    case 2 : a+=k[0];
399             break;
400    case 1 : a+=k8[0];
401             break;
402    case 0 : return c;                     /* zero length requires no mixing */
403    }
404
405  } else {                        /* need to read the key one byte at a time */
406    const uint8_t *k = (const uint8_t *)key;
407
408    /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
409    while (length > 12)
410    {
411      a += k[0];
412      a += ((uint32_t)k[1])<<8;
413      a += ((uint32_t)k[2])<<16;
414      a += ((uint32_t)k[3])<<24;
415      b += k[4];
416      b += ((uint32_t)k[5])<<8;
417      b += ((uint32_t)k[6])<<16;
418      b += ((uint32_t)k[7])<<24;
419      c += k[8];
420      c += ((uint32_t)k[9])<<8;
421      c += ((uint32_t)k[10])<<16;
422      c += ((uint32_t)k[11])<<24;
423      mix(a,b,c);
424      length -= 12;
425      k += 12;
426    }
427
428    /*-------------------------------- last block: affect all 32 bits of (c) */
429    switch(length)                   /* all the case statements fall through */
430    {
431    case 12: c+=((uint32_t)k[11])<<24;
432    case 11: c+=((uint32_t)k[10])<<16;
433    case 10: c+=((uint32_t)k[9])<<8;
434    case 9 : c+=k[8];
435    case 8 : b+=((uint32_t)k[7])<<24;
436    case 7 : b+=((uint32_t)k[6])<<16;
437    case 6 : b+=((uint32_t)k[5])<<8;
438    case 5 : b+=k[4];
439    case 4 : a+=((uint32_t)k[3])<<24;
440    case 3 : a+=((uint32_t)k[2])<<16;
441    case 2 : a+=((uint32_t)k[1])<<8;
442    case 1 : a+=k[0];
443             break;
444    case 0 : return c;
445    }
446  }
447
448  final(a,b,c);
449  return c;
450}
451
452
453/*
454 * hashlittle2: return 2 32-bit hash values
455 *
456 * This is identical to hashlittle(), except it returns two 32-bit hash
457 * values instead of just one.  This is good enough for hash table
458 * lookup with 2^^64 buckets, or if you want a second hash if you're not
459 * happy with the first, or if you want a probably-unique 64-bit ID for
460 * the key.  *pc is better mixed than *pb, so use *pc first.  If you want
461 * a 64-bit value do something like "*pc + (((uint64_t)*pb)<<32)".
462 */
463void hashlittle2(
464  const void *key,       /* the key to hash */
465  size_t      length,    /* length of the key */
466  uint32_t   *pc,        /* IN: primary initval, OUT: primary hash */
467  uint32_t   *pb)        /* IN: secondary initval, OUT: secondary hash */
468{
469  uint32_t a,b,c;                                          /* internal state */
470  union { const void *ptr; size_t i; } u;     /* needed for Mac Powerbook G4 */
471
472  /* Set up the internal state */
473  a = b = c = 0xdeadbeef + ((uint32_t)length) + *pc;
474  c += *pb;
475
476  u.ptr = key;
477  if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
478    const uint32_t *k = (const uint32_t *)key;         /* read 32-bit chunks */
479    const uint8_t  *k8;
480
481    /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
482    while (length > 12)
483    {
484      a += k[0];
485      b += k[1];
486      c += k[2];
487      mix(a,b,c);
488      length -= 12;
489      k += 3;
490    }
491
492    /*----------------------------- handle the last (probably partial) block */
493    /*
494     * "k[2]&0xffffff" actually reads beyond the end of the string, but
495     * then masks off the part it's not allowed to read.  Because the
496     * string is aligned, the masked-off tail is in the same word as the
497     * rest of the string.  Every machine with memory protection I've seen
498     * does it on word boundaries, so is OK with this.  But VALGRIND will
499     * still catch it and complain.  The masking trick does make the hash
500     * noticably faster for short strings (like English words).
501     */
502#ifndef VALGRIND
503
504    switch(length)
505    {
506    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
507    case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
508    case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
509    case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
510    case 8 : b+=k[1]; a+=k[0]; break;
511    case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
512    case 6 : b+=k[1]&0xffff; a+=k[0]; break;
513    case 5 : b+=k[1]&0xff; a+=k[0]; break;
514    case 4 : a+=k[0]; break;
515    case 3 : a+=k[0]&0xffffff; break;
516    case 2 : a+=k[0]&0xffff; break;
517    case 1 : a+=k[0]&0xff; break;
518    case 0 : *pc=c; *pb=b; return;  /* zero length strings require no mixing */
519    }
520
521#else /* make valgrind happy */
522
523    k8 = (const uint8_t *)k;
524    switch(length)
525    {
526    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
527    case 11: c+=((uint32_t)k8[10])<<16;  /* fall through */
528    case 10: c+=((uint32_t)k8[9])<<8;    /* fall through */
529    case 9 : c+=k8[8];                   /* fall through */
530    case 8 : b+=k[1]; a+=k[0]; break;
531    case 7 : b+=((uint32_t)k8[6])<<16;   /* fall through */
532    case 6 : b+=((uint32_t)k8[5])<<8;    /* fall through */
533    case 5 : b+=k8[4];                   /* fall through */
534    case 4 : a+=k[0]; break;
535    case 3 : a+=((uint32_t)k8[2])<<16;   /* fall through */
536    case 2 : a+=((uint32_t)k8[1])<<8;    /* fall through */
537    case 1 : a+=k8[0]; break;
538    case 0 : *pc=c; *pb=b; return;  /* zero length strings require no mixing */
539    }
540
541#endif /* !valgrind */
542
543  } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
544    const uint16_t *k = (const uint16_t *)key;         /* read 16-bit chunks */
545    const uint8_t  *k8;
546
547    /*--------------- all but last block: aligned reads and different mixing */
548    while (length > 12)
549    {
550      a += k[0] + (((uint32_t)k[1])<<16);
551      b += k[2] + (((uint32_t)k[3])<<16);
552      c += k[4] + (((uint32_t)k[5])<<16);
553      mix(a,b,c);
554      length -= 12;
555      k += 6;
556    }
557
558    /*----------------------------- handle the last (probably partial) block */
559    k8 = (const uint8_t *)k;
560    switch(length)
561    {
562    case 12: c+=k[4]+(((uint32_t)k[5])<<16);
563             b+=k[2]+(((uint32_t)k[3])<<16);
564             a+=k[0]+(((uint32_t)k[1])<<16);
565             break;
566    case 11: c+=((uint32_t)k8[10])<<16;     /* fall through */
567    case 10: c+=k[4];
568             b+=k[2]+(((uint32_t)k[3])<<16);
569             a+=k[0]+(((uint32_t)k[1])<<16);
570             break;
571    case 9 : c+=k8[8];                      /* fall through */
572    case 8 : b+=k[2]+(((uint32_t)k[3])<<16);
573             a+=k[0]+(((uint32_t)k[1])<<16);
574             break;
575    case 7 : b+=((uint32_t)k8[6])<<16;      /* fall through */
576    case 6 : b+=k[2];
577             a+=k[0]+(((uint32_t)k[1])<<16);
578             break;
579    case 5 : b+=k8[4];                      /* fall through */
580    case 4 : a+=k[0]+(((uint32_t)k[1])<<16);
581             break;
582    case 3 : a+=((uint32_t)k8[2])<<16;      /* fall through */
583    case 2 : a+=k[0];
584             break;
585    case 1 : a+=k8[0];
586             break;
587    case 0 : *pc=c; *pb=b; return;  /* zero length strings require no mixing */
588    }
589
590  } else {                        /* need to read the key one byte at a time */
591    const uint8_t *k = (const uint8_t *)key;
592
593    /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
594    while (length > 12)
595    {
596      a += k[0];
597      a += ((uint32_t)k[1])<<8;
598      a += ((uint32_t)k[2])<<16;
599      a += ((uint32_t)k[3])<<24;
600      b += k[4];
601      b += ((uint32_t)k[5])<<8;
602      b += ((uint32_t)k[6])<<16;
603      b += ((uint32_t)k[7])<<24;
604      c += k[8];
605      c += ((uint32_t)k[9])<<8;
606      c += ((uint32_t)k[10])<<16;
607      c += ((uint32_t)k[11])<<24;
608      mix(a,b,c);
609      length -= 12;
610      k += 12;
611    }
612
613    /*-------------------------------- last block: affect all 32 bits of (c) */
614    switch(length)                   /* all the case statements fall through */
615    {
616    case 12: c+=((uint32_t)k[11])<<24;
617    case 11: c+=((uint32_t)k[10])<<16;
618    case 10: c+=((uint32_t)k[9])<<8;
619    case 9 : c+=k[8];
620    case 8 : b+=((uint32_t)k[7])<<24;
621    case 7 : b+=((uint32_t)k[6])<<16;
622    case 6 : b+=((uint32_t)k[5])<<8;
623    case 5 : b+=k[4];
624    case 4 : a+=((uint32_t)k[3])<<24;
625    case 3 : a+=((uint32_t)k[2])<<16;
626    case 2 : a+=((uint32_t)k[1])<<8;
627    case 1 : a+=k[0];
628             break;
629    case 0 : *pc=c; *pb=b; return;  /* zero length strings require no mixing */
630    }
631  }
632
633  final(a,b,c);
634  *pc=c; *pb=b;
635}
636
637
638
639/*
640 * hashbig():
641 * This is the same as hashword() on big-endian machines.  It is different
642 * from hashlittle() on all machines.  hashbig() takes advantage of
643 * big-endian byte ordering.
644 */
645uint32_t hashbig( const void *key, size_t length, uint32_t initval)
646{
647  uint32_t a,b,c;
648  union { const void *ptr; size_t i; } u; /* to cast key to (size_t) happily */
649
650  /* Set up the internal state */
651  a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
652
653  u.ptr = key;
654  if (HASH_BIG_ENDIAN && ((u.i & 0x3) == 0)) {
655    const uint32_t *k = (const uint32_t *)key;         /* read 32-bit chunks */
656    const uint8_t  *k8;
657
658    /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
659    while (length > 12)
660    {
661      a += k[0];
662      b += k[1];
663      c += k[2];
664      mix(a,b,c);
665      length -= 12;
666      k += 3;
667    }
668
669    /*----------------------------- handle the last (probably partial) block */
670    /*
671     * "k[2]<<8" actually reads beyond the end of the string, but
672     * then shifts out the part it's not allowed to read.  Because the
673     * string is aligned, the illegal read is in the same word as the
674     * rest of the string.  Every machine with memory protection I've seen
675     * does it on word boundaries, so is OK with this.  But VALGRIND will
676     * still catch it and complain.  The masking trick does make the hash
677     * noticably faster for short strings (like English words).
678     */
679#ifndef VALGRIND
680
681    switch(length)
682    {
683    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
684    case 11: c+=k[2]&0xffffff00; b+=k[1]; a+=k[0]; break;
685    case 10: c+=k[2]&0xffff0000; b+=k[1]; a+=k[0]; break;
686    case 9 : c+=k[2]&0xff000000; b+=k[1]; a+=k[0]; break;
687    case 8 : b+=k[1]; a+=k[0]; break;
688    case 7 : b+=k[1]&0xffffff00; a+=k[0]; break;
689    case 6 : b+=k[1]&0xffff0000; a+=k[0]; break;
690    case 5 : b+=k[1]&0xff000000; a+=k[0]; break;
691    case 4 : a+=k[0]; break;
692    case 3 : a+=k[0]&0xffffff00; break;
693    case 2 : a+=k[0]&0xffff0000; break;
694    case 1 : a+=k[0]&0xff000000; break;
695    case 0 : return c;              /* zero length strings require no mixing */
696    }
697
698#else  /* make valgrind happy */
699
700    k8 = (const uint8_t *)k;
701    switch(length)                   /* all the case statements fall through */
702    {
703    case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
704    case 11: c+=((uint32_t)k8[10])<<8;  /* fall through */
705    case 10: c+=((uint32_t)k8[9])<<16;  /* fall through */
706    case 9 : c+=((uint32_t)k8[8])<<24;  /* fall through */
707    case 8 : b+=k[1]; a+=k[0]; break;
708    case 7 : b+=((uint32_t)k8[6])<<8;   /* fall through */
709    case 6 : b+=((uint32_t)k8[5])<<16;  /* fall through */
710    case 5 : b+=((uint32_t)k8[4])<<24;  /* fall through */
711    case 4 : a+=k[0]; break;
712    case 3 : a+=((uint32_t)k8[2])<<8;   /* fall through */
713    case 2 : a+=((uint32_t)k8[1])<<16;  /* fall through */
714    case 1 : a+=((uint32_t)k8[0])<<24; break;
715    case 0 : return c;
716    }
717
718#endif /* !VALGRIND */
719
720  } else {                        /* need to read the key one byte at a time */
721    const uint8_t *k = (const uint8_t *)key;
722
723    /*--------------- all but the last block: affect some 32 bits of (a,b,c) */
724    while (length > 12)
725    {
726      a += ((uint32_t)k[0])<<24;
727      a += ((uint32_t)k[1])<<16;
728      a += ((uint32_t)k[2])<<8;
729      a += ((uint32_t)k[3]);
730      b += ((uint32_t)k[4])<<24;
731      b += ((uint32_t)k[5])<<16;
732      b += ((uint32_t)k[6])<<8;
733      b += ((uint32_t)k[7]);
734      c += ((uint32_t)k[8])<<24;
735      c += ((uint32_t)k[9])<<16;
736      c += ((uint32_t)k[10])<<8;
737      c += ((uint32_t)k[11]);
738      mix(a,b,c);
739      length -= 12;
740      k += 12;
741    }
742
743    /*-------------------------------- last block: affect all 32 bits of (c) */
744    switch(length)                   /* all the case statements fall through */
745    {
746    case 12: c+=k[11];
747    case 11: c+=((uint32_t)k[10])<<8;
748    case 10: c+=((uint32_t)k[9])<<16;
749    case 9 : c+=((uint32_t)k[8])<<24;
750    case 8 : b+=k[7];
751    case 7 : b+=((uint32_t)k[6])<<8;
752    case 6 : b+=((uint32_t)k[5])<<16;
753    case 5 : b+=((uint32_t)k[4])<<24;
754    case 4 : a+=k[3];
755    case 3 : a+=((uint32_t)k[2])<<8;
756    case 2 : a+=((uint32_t)k[1])<<16;
757    case 1 : a+=((uint32_t)k[0])<<24;
758             break;
759    case 0 : return c;
760    }
761  }
762
763  final(a,b,c);
764  return c;
765}
766
767
768#ifdef SELF_TEST
769
770/* used for timings */
771void driver1()
772{
773  uint8_t buf[256];
774  uint32_t i;
775  uint32_t h=0;
776  time_t a,z;
777
778  time(&a);
779  for (i=0; i<256; ++i) buf[i] = 'x';
780  for (i=0; i<1; ++i)
781  {
782    h = hashlittle(&buf[0],1,h);
783  }
784  time(&z);
785  if (z-a > 0) printf("time %d %.8x\n", z-a, h);
786}
787
788/* check that every input bit changes every output bit half the time */
789#define HASHSTATE 1
790#define HASHLEN   1
791#define MAXPAIR 60
792#define MAXLEN  70
793void driver2()
794{
795  uint8_t qa[MAXLEN+1], qb[MAXLEN+2], *a = &qa[0], *b = &qb[1];
796  uint32_t c[HASHSTATE], d[HASHSTATE], i=0, j=0, k, l, m=0, z;
797  uint32_t e[HASHSTATE],f[HASHSTATE],g[HASHSTATE],h[HASHSTATE];
798  uint32_t x[HASHSTATE],y[HASHSTATE];
799  uint32_t hlen;
800
801  printf("No more than %d trials should ever be needed \n",MAXPAIR/2);
802  for (hlen=0; hlen < MAXLEN; ++hlen)
803  {
804    z=0;
805    for (i=0; i<hlen; ++i)  /*----------------------- for each input byte, */
806    {
807      for (j=0; j<8; ++j)   /*------------------------ for each input bit, */
808      {
809	for (m=1; m<8; ++m) /*------------ for serveral possible initvals, */
810	{
811	  for (l=0; l<HASHSTATE; ++l)
812	    e[l]=f[l]=g[l]=h[l]=x[l]=y[l]=~((uint32_t)0);
813
814      	  /*---- check that every output bit is affected by that input bit */
815	  for (k=0; k<MAXPAIR; k+=2)
816	  {
817	    uint32_t finished=1;
818	    /* keys have one bit different */
819	    for (l=0; l<hlen+1; ++l) {a[l] = b[l] = (uint8_t)0;}
820	    /* have a and b be two keys differing in only one bit */
821	    a[i] ^= (k<<j);
822	    a[i] ^= (k>>(8-j));
823	     c[0] = hashlittle(a, hlen, m);
824	    b[i] ^= ((k+1)<<j);
825	    b[i] ^= ((k+1)>>(8-j));
826	     d[0] = hashlittle(b, hlen, m);
827	    /* check every bit is 1, 0, set, and not set at least once */
828	    for (l=0; l<HASHSTATE; ++l)
829	    {
830	      e[l] &= (c[l]^d[l]);
831	      f[l] &= ~(c[l]^d[l]);
832	      g[l] &= c[l];
833	      h[l] &= ~c[l];
834	      x[l] &= d[l];
835	      y[l] &= ~d[l];
836	      if (e[l]|f[l]|g[l]|h[l]|x[l]|y[l]) finished=0;
837	    }
838	    if (finished) break;
839	  }
840	  if (k>z) z=k;
841	  if (k==MAXPAIR)
842	  {
843	     printf("Some bit didn't change: ");
844	     printf("%.8x %.8x %.8x %.8x %.8x %.8x  ",
845	            e[0],f[0],g[0],h[0],x[0],y[0]);
846	     printf("i %d j %d m %d len %d\n", i, j, m, hlen);
847	  }
848	  if (z==MAXPAIR) goto done;
849	}
850      }
851    }
852   done:
853    if (z < MAXPAIR)
854    {
855      printf("Mix success  %2d bytes  %2d initvals  ",i,m);
856      printf("required  %d  trials\n", z/2);
857    }
858  }
859  printf("\n");
860}
861
862/* Check for reading beyond the end of the buffer and alignment problems */
863void driver3()
864{
865  uint8_t buf[MAXLEN+20], *b;
866  uint32_t len;
867  uint8_t q[] = "This is the time for all good men to come to the aid of their country...";
868  uint32_t h;
869  uint8_t qq[] = "xThis is the time for all good men to come to the aid of their country...";
870  uint32_t i;
871  uint8_t qqq[] = "xxThis is the time for all good men to come to the aid of their country...";
872  uint32_t j;
873  uint8_t qqqq[] = "xxxThis is the time for all good men to come to the aid of their country...";
874  uint32_t ref,x,y;
875  uint8_t *p;
876
877  printf("Endianness.  These lines should all be the same (for values filled in):\n");
878  printf("%.8x                            %.8x                            %.8x\n",
879         hashword((const uint32_t *)q, (sizeof(q)-1)/4, 13),
880         hashword((const uint32_t *)q, (sizeof(q)-5)/4, 13),
881         hashword((const uint32_t *)q, (sizeof(q)-9)/4, 13));
882  p = q;
883  printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
884         hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
885         hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
886         hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
887         hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
888         hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
889         hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
890  p = &qq[1];
891  printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
892         hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
893         hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
894         hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
895         hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
896         hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
897         hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
898  p = &qqq[2];
899  printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
900         hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
901         hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
902         hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
903         hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
904         hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
905         hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
906  p = &qqqq[3];
907  printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n",
908         hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13),
909         hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13),
910         hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13),
911         hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13),
912         hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13),
913         hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13));
914  printf("\n");
915
916  /* check that hashlittle2 and hashlittle produce the same results */
917  i=47; j=0;
918  hashlittle2(q, sizeof(q), &i, &j);
919  if (hashlittle(q, sizeof(q), 47) != i)
920    printf("hashlittle2 and hashlittle mismatch\n");
921
922  /* check that hashword2 and hashword produce the same results */
923  len = 0xdeadbeef;
924  i=47, j=0;
925  hashword2(&len, 1, &i, &j);
926  if (hashword(&len, 1, 47) != i)
927    printf("hashword2 and hashword mismatch %x %x\n",
928	   i, hashword(&len, 1, 47));
929
930  /* check hashlittle doesn't read before or after the ends of the string */
931  for (h=0, b=buf+1; h<8; ++h, ++b)
932  {
933    for (i=0; i<MAXLEN; ++i)
934    {
935      len = i;
936      for (j=0; j<i; ++j) *(b+j)=0;
937
938      /* these should all be equal */
939      ref = hashlittle(b, len, (uint32_t)1);
940      *(b+i)=(uint8_t)~0;
941      *(b-1)=(uint8_t)~0;
942      x = hashlittle(b, len, (uint32_t)1);
943      y = hashlittle(b, len, (uint32_t)1);
944      if ((ref != x) || (ref != y))
945      {
946	printf("alignment error: %.8x %.8x %.8x %d %d\n",ref,x,y,
947               h, i);
948      }
949    }
950  }
951}
952
953/* check for problems with nulls */
954 void driver4()
955{
956  uint8_t buf[1];
957  uint32_t h,i,state[HASHSTATE];
958
959
960  buf[0] = ~0;
961  for (i=0; i<HASHSTATE; ++i) state[i] = 1;
962  printf("These should all be different\n");
963  for (i=0, h=0; i<8; ++i)
964  {
965    h = hashlittle(buf, 0, h);
966    printf("%2ld  0-byte strings, hash is  %.8x\n", i, h);
967  }
968}
969
970void driver5()
971{
972  uint32_t b,c;
973  b=0, c=0, hashlittle2("", 0, &c, &b);
974  printf("hash is %.8lx %.8lx\n", c, b);   /* deadbeef deadbeef */
975  b=0xdeadbeef, c=0, hashlittle2("", 0, &c, &b);
976  printf("hash is %.8lx %.8lx\n", c, b);   /* bd5b7dde deadbeef */
977  b=0xdeadbeef, c=0xdeadbeef, hashlittle2("", 0, &c, &b);
978  printf("hash is %.8lx %.8lx\n", c, b);   /* 9c093ccd bd5b7dde */
979  b=0, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b);
980  printf("hash is %.8lx %.8lx\n", c, b);   /* 17770551 ce7226e6 */
981  b=1, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b);
982  printf("hash is %.8lx %.8lx\n", c, b);   /* e3607cae bd371de4 */
983  b=0, c=1, hashlittle2("Four score and seven years ago", 30, &c, &b);
984  printf("hash is %.8lx %.8lx\n", c, b);   /* cd628161 6cbea4b3 */
985  c = hashlittle("Four score and seven years ago", 30, 0);
986  printf("hash is %.8lx\n", c);   /* 17770551 */
987  c = hashlittle("Four score and seven years ago", 30, 1);
988  printf("hash is %.8lx\n", c);   /* cd628161 */
989}
990
991
992int main()
993{
994  driver1();   /* test that the key is hashed: used for timings */
995  driver2();   /* test that whole key is hashed thoroughly */
996  driver3();   /* test that nothing but the key is hashed */
997  driver4();   /* test hashing multiple buffers (all buffers are null) */
998  driver5();   /* test the hash against known vectors */
999  return 1;
1000}
1001
1002#endif  /* SELF_TEST */
1003