1 |
274 |
jeremybenn |
/* An expandable hash tables datatype.
|
2 |
|
|
Copyright (C) 1999, 2000, 2001, 2002, 2003, 2004, 2009
|
3 |
|
|
Free Software Foundation, Inc.
|
4 |
|
|
Contributed by Vladimir Makarov (vmakarov@cygnus.com).
|
5 |
|
|
|
6 |
|
|
This file is part of the libiberty library.
|
7 |
|
|
Libiberty is free software; you can redistribute it and/or
|
8 |
|
|
modify it under the terms of the GNU Library General Public
|
9 |
|
|
License as published by the Free Software Foundation; either
|
10 |
|
|
version 2 of the License, or (at your option) any later version.
|
11 |
|
|
|
12 |
|
|
Libiberty is distributed in the hope that it will be useful,
|
13 |
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
14 |
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
15 |
|
|
Library General Public License for more details.
|
16 |
|
|
|
17 |
|
|
You should have received a copy of the GNU Library General Public
|
18 |
|
|
License along with libiberty; see the file COPYING.LIB. If
|
19 |
|
|
not, write to the Free Software Foundation, Inc., 51 Franklin Street - Fifth Floor,
|
20 |
|
|
Boston, MA 02110-1301, USA. */
|
21 |
|
|
|
22 |
|
|
/* This package implements basic hash table functionality. It is possible
|
23 |
|
|
to search for an entry, create an entry and destroy an entry.
|
24 |
|
|
|
25 |
|
|
Elements in the table are generic pointers.
|
26 |
|
|
|
27 |
|
|
The size of the table is not fixed; if the occupancy of the table
|
28 |
|
|
grows too high the hash table will be expanded.
|
29 |
|
|
|
30 |
|
|
The abstract data implementation is based on generalized Algorithm D
|
31 |
|
|
from Knuth's book "The art of computer programming". Hash table is
|
32 |
|
|
expanded by creation of new hash table and transferring elements from
|
33 |
|
|
the old table to the new table. */
|
34 |
|
|
|
35 |
|
|
#ifdef HAVE_CONFIG_H
|
36 |
|
|
#include "config.h"
|
37 |
|
|
#endif
|
38 |
|
|
|
39 |
|
|
#include <sys/types.h>
|
40 |
|
|
|
41 |
|
|
#ifdef HAVE_STDLIB_H
|
42 |
|
|
#include <stdlib.h>
|
43 |
|
|
#endif
|
44 |
|
|
#ifdef HAVE_STRING_H
|
45 |
|
|
#include <string.h>
|
46 |
|
|
#endif
|
47 |
|
|
#ifdef HAVE_MALLOC_H
|
48 |
|
|
#include <malloc.h>
|
49 |
|
|
#endif
|
50 |
|
|
#ifdef HAVE_LIMITS_H
|
51 |
|
|
#include <limits.h>
|
52 |
|
|
#endif
|
53 |
|
|
#ifdef HAVE_INTTYPES_H
|
54 |
|
|
#include <inttypes.h>
|
55 |
|
|
#endif
|
56 |
|
|
#ifdef HAVE_STDINT_H
|
57 |
|
|
#include <stdint.h>
|
58 |
|
|
#endif
|
59 |
|
|
|
60 |
|
|
#include <stdio.h>
|
61 |
|
|
|
62 |
|
|
#include "libiberty.h"
|
63 |
|
|
#include "ansidecl.h"
|
64 |
|
|
#include "hashtab.h"
|
65 |
|
|
|
66 |
|
|
#ifndef CHAR_BIT
|
67 |
|
|
#define CHAR_BIT 8
|
68 |
|
|
#endif
|
69 |
|
|
|
70 |
|
|
static unsigned int higher_prime_index (unsigned long);
|
71 |
|
|
static hashval_t htab_mod_1 (hashval_t, hashval_t, hashval_t, int);
|
72 |
|
|
static hashval_t htab_mod (hashval_t, htab_t);
|
73 |
|
|
static hashval_t htab_mod_m2 (hashval_t, htab_t);
|
74 |
|
|
static hashval_t hash_pointer (const void *);
|
75 |
|
|
static int eq_pointer (const void *, const void *);
|
76 |
|
|
static int htab_expand (htab_t);
|
77 |
|
|
static PTR *find_empty_slot_for_expand (htab_t, hashval_t);
|
78 |
|
|
|
79 |
|
|
/* At some point, we could make these be NULL, and modify the
|
80 |
|
|
hash-table routines to handle NULL specially; that would avoid
|
81 |
|
|
function-call overhead for the common case of hashing pointers. */
|
82 |
|
|
htab_hash htab_hash_pointer = hash_pointer;
|
83 |
|
|
htab_eq htab_eq_pointer = eq_pointer;
|
84 |
|
|
|
85 |
|
|
/* Table of primes and multiplicative inverses.
|
86 |
|
|
|
87 |
|
|
Note that these are not minimally reduced inverses. Unlike when generating
|
88 |
|
|
code to divide by a constant, we want to be able to use the same algorithm
|
89 |
|
|
all the time. All of these inverses (are implied to) have bit 32 set.
|
90 |
|
|
|
91 |
|
|
For the record, here's the function that computed the table; it's a
|
92 |
|
|
vastly simplified version of the function of the same name from gcc. */
|
93 |
|
|
|
94 |
|
|
#if 0
|
95 |
|
|
unsigned int
|
96 |
|
|
ceil_log2 (unsigned int x)
|
97 |
|
|
{
|
98 |
|
|
int i;
|
99 |
|
|
for (i = 31; i >= 0 ; --i)
|
100 |
|
|
if (x > (1u << i))
|
101 |
|
|
return i+1;
|
102 |
|
|
abort ();
|
103 |
|
|
}
|
104 |
|
|
|
105 |
|
|
unsigned int
|
106 |
|
|
choose_multiplier (unsigned int d, unsigned int *mlp, unsigned char *shiftp)
|
107 |
|
|
{
|
108 |
|
|
unsigned long long mhigh;
|
109 |
|
|
double nx;
|
110 |
|
|
int lgup, post_shift;
|
111 |
|
|
int pow, pow2;
|
112 |
|
|
int n = 32, precision = 32;
|
113 |
|
|
|
114 |
|
|
lgup = ceil_log2 (d);
|
115 |
|
|
pow = n + lgup;
|
116 |
|
|
pow2 = n + lgup - precision;
|
117 |
|
|
|
118 |
|
|
nx = ldexp (1.0, pow) + ldexp (1.0, pow2);
|
119 |
|
|
mhigh = nx / d;
|
120 |
|
|
|
121 |
|
|
*shiftp = lgup - 1;
|
122 |
|
|
*mlp = mhigh;
|
123 |
|
|
return mhigh >> 32;
|
124 |
|
|
}
|
125 |
|
|
#endif
|
126 |
|
|
|
127 |
|
|
struct prime_ent
|
128 |
|
|
{
|
129 |
|
|
hashval_t prime;
|
130 |
|
|
hashval_t inv;
|
131 |
|
|
hashval_t inv_m2; /* inverse of prime-2 */
|
132 |
|
|
hashval_t shift;
|
133 |
|
|
};
|
134 |
|
|
|
135 |
|
|
static struct prime_ent const prime_tab[] = {
|
136 |
|
|
{ 7, 0x24924925, 0x9999999b, 2 },
|
137 |
|
|
{ 13, 0x3b13b13c, 0x745d1747, 3 },
|
138 |
|
|
{ 31, 0x08421085, 0x1a7b9612, 4 },
|
139 |
|
|
{ 61, 0x0c9714fc, 0x15b1e5f8, 5 },
|
140 |
|
|
{ 127, 0x02040811, 0x0624dd30, 6 },
|
141 |
|
|
{ 251, 0x05197f7e, 0x073260a5, 7 },
|
142 |
|
|
{ 509, 0x01824366, 0x02864fc8, 8 },
|
143 |
|
|
{ 1021, 0x00c0906d, 0x014191f7, 9 },
|
144 |
|
|
{ 2039, 0x0121456f, 0x0161e69e, 10 },
|
145 |
|
|
{ 4093, 0x00300902, 0x00501908, 11 },
|
146 |
|
|
{ 8191, 0x00080041, 0x00180241, 12 },
|
147 |
|
|
{ 16381, 0x000c0091, 0x00140191, 13 },
|
148 |
|
|
{ 32749, 0x002605a5, 0x002a06e6, 14 },
|
149 |
|
|
{ 65521, 0x000f00e2, 0x00110122, 15 },
|
150 |
|
|
{ 131071, 0x00008001, 0x00018003, 16 },
|
151 |
|
|
{ 262139, 0x00014002, 0x0001c004, 17 },
|
152 |
|
|
{ 524287, 0x00002001, 0x00006001, 18 },
|
153 |
|
|
{ 1048573, 0x00003001, 0x00005001, 19 },
|
154 |
|
|
{ 2097143, 0x00004801, 0x00005801, 20 },
|
155 |
|
|
{ 4194301, 0x00000c01, 0x00001401, 21 },
|
156 |
|
|
{ 8388593, 0x00001e01, 0x00002201, 22 },
|
157 |
|
|
{ 16777213, 0x00000301, 0x00000501, 23 },
|
158 |
|
|
{ 33554393, 0x00001381, 0x00001481, 24 },
|
159 |
|
|
{ 67108859, 0x00000141, 0x000001c1, 25 },
|
160 |
|
|
{ 134217689, 0x000004e1, 0x00000521, 26 },
|
161 |
|
|
{ 268435399, 0x00000391, 0x000003b1, 27 },
|
162 |
|
|
{ 536870909, 0x00000019, 0x00000029, 28 },
|
163 |
|
|
{ 1073741789, 0x0000008d, 0x00000095, 29 },
|
164 |
|
|
{ 2147483647, 0x00000003, 0x00000007, 30 },
|
165 |
|
|
/* Avoid "decimal constant so large it is unsigned" for 4294967291. */
|
166 |
|
|
{ 0xfffffffb, 0x00000006, 0x00000008, 31 }
|
167 |
|
|
};
|
168 |
|
|
|
169 |
|
|
/* The following function returns an index into the above table of the
|
170 |
|
|
nearest prime number which is greater than N, and near a power of two. */
|
171 |
|
|
|
172 |
|
|
static unsigned int
|
173 |
|
|
higher_prime_index (unsigned long n)
|
174 |
|
|
{
|
175 |
|
|
unsigned int low = 0;
|
176 |
|
|
unsigned int high = sizeof(prime_tab) / sizeof(prime_tab[0]);
|
177 |
|
|
|
178 |
|
|
while (low != high)
|
179 |
|
|
{
|
180 |
|
|
unsigned int mid = low + (high - low) / 2;
|
181 |
|
|
if (n > prime_tab[mid].prime)
|
182 |
|
|
low = mid + 1;
|
183 |
|
|
else
|
184 |
|
|
high = mid;
|
185 |
|
|
}
|
186 |
|
|
|
187 |
|
|
/* If we've run out of primes, abort. */
|
188 |
|
|
if (n > prime_tab[low].prime)
|
189 |
|
|
{
|
190 |
|
|
fprintf (stderr, "Cannot find prime bigger than %lu\n", n);
|
191 |
|
|
abort ();
|
192 |
|
|
}
|
193 |
|
|
|
194 |
|
|
return low;
|
195 |
|
|
}
|
196 |
|
|
|
197 |
|
|
/* Returns a hash code for P. */
|
198 |
|
|
|
199 |
|
|
static hashval_t
|
200 |
|
|
hash_pointer (const PTR p)
|
201 |
|
|
{
|
202 |
|
|
return (hashval_t) ((intptr_t)p >> 3);
|
203 |
|
|
}
|
204 |
|
|
|
205 |
|
|
/* Returns non-zero if P1 and P2 are equal. */
|
206 |
|
|
|
207 |
|
|
static int
|
208 |
|
|
eq_pointer (const PTR p1, const PTR p2)
|
209 |
|
|
{
|
210 |
|
|
return p1 == p2;
|
211 |
|
|
}
|
212 |
|
|
|
213 |
|
|
|
214 |
|
|
/* The parens around the function names in the next two definitions
|
215 |
|
|
are essential in order to prevent macro expansions of the name.
|
216 |
|
|
The bodies, however, are expanded as expected, so they are not
|
217 |
|
|
recursive definitions. */
|
218 |
|
|
|
219 |
|
|
/* Return the current size of given hash table. */
|
220 |
|
|
|
221 |
|
|
#define htab_size(htab) ((htab)->size)
|
222 |
|
|
|
223 |
|
|
size_t
|
224 |
|
|
(htab_size) (htab_t htab)
|
225 |
|
|
{
|
226 |
|
|
return htab_size (htab);
|
227 |
|
|
}
|
228 |
|
|
|
229 |
|
|
/* Return the current number of elements in given hash table. */
|
230 |
|
|
|
231 |
|
|
#define htab_elements(htab) ((htab)->n_elements - (htab)->n_deleted)
|
232 |
|
|
|
233 |
|
|
size_t
|
234 |
|
|
(htab_elements) (htab_t htab)
|
235 |
|
|
{
|
236 |
|
|
return htab_elements (htab);
|
237 |
|
|
}
|
238 |
|
|
|
239 |
|
|
/* Return X % Y. */
|
240 |
|
|
|
241 |
|
|
static inline hashval_t
|
242 |
|
|
htab_mod_1 (hashval_t x, hashval_t y, hashval_t inv, int shift)
|
243 |
|
|
{
|
244 |
|
|
/* The multiplicative inverses computed above are for 32-bit types, and
|
245 |
|
|
requires that we be able to compute a highpart multiply. */
|
246 |
|
|
#ifdef UNSIGNED_64BIT_TYPE
|
247 |
|
|
__extension__ typedef UNSIGNED_64BIT_TYPE ull;
|
248 |
|
|
if (sizeof (hashval_t) * CHAR_BIT <= 32)
|
249 |
|
|
{
|
250 |
|
|
hashval_t t1, t2, t3, t4, q, r;
|
251 |
|
|
|
252 |
|
|
t1 = ((ull)x * inv) >> 32;
|
253 |
|
|
t2 = x - t1;
|
254 |
|
|
t3 = t2 >> 1;
|
255 |
|
|
t4 = t1 + t3;
|
256 |
|
|
q = t4 >> shift;
|
257 |
|
|
r = x - (q * y);
|
258 |
|
|
|
259 |
|
|
return r;
|
260 |
|
|
}
|
261 |
|
|
#endif
|
262 |
|
|
|
263 |
|
|
/* Otherwise just use the native division routines. */
|
264 |
|
|
return x % y;
|
265 |
|
|
}
|
266 |
|
|
|
267 |
|
|
/* Compute the primary hash for HASH given HTAB's current size. */
|
268 |
|
|
|
269 |
|
|
static inline hashval_t
|
270 |
|
|
htab_mod (hashval_t hash, htab_t htab)
|
271 |
|
|
{
|
272 |
|
|
const struct prime_ent *p = &prime_tab[htab->size_prime_index];
|
273 |
|
|
return htab_mod_1 (hash, p->prime, p->inv, p->shift);
|
274 |
|
|
}
|
275 |
|
|
|
276 |
|
|
/* Compute the secondary hash for HASH given HTAB's current size. */
|
277 |
|
|
|
278 |
|
|
static inline hashval_t
|
279 |
|
|
htab_mod_m2 (hashval_t hash, htab_t htab)
|
280 |
|
|
{
|
281 |
|
|
const struct prime_ent *p = &prime_tab[htab->size_prime_index];
|
282 |
|
|
return 1 + htab_mod_1 (hash, p->prime - 2, p->inv_m2, p->shift);
|
283 |
|
|
}
|
284 |
|
|
|
285 |
|
|
/* This function creates table with length slightly longer than given
|
286 |
|
|
source length. Created hash table is initiated as empty (all the
|
287 |
|
|
hash table entries are HTAB_EMPTY_ENTRY). The function returns the
|
288 |
|
|
created hash table, or NULL if memory allocation fails. */
|
289 |
|
|
|
290 |
|
|
htab_t
|
291 |
|
|
htab_create_alloc (size_t size, htab_hash hash_f, htab_eq eq_f,
|
292 |
|
|
htab_del del_f, htab_alloc alloc_f, htab_free free_f)
|
293 |
|
|
{
|
294 |
|
|
htab_t result;
|
295 |
|
|
unsigned int size_prime_index;
|
296 |
|
|
|
297 |
|
|
size_prime_index = higher_prime_index (size);
|
298 |
|
|
size = prime_tab[size_prime_index].prime;
|
299 |
|
|
|
300 |
|
|
result = (htab_t) (*alloc_f) (1, sizeof (struct htab));
|
301 |
|
|
if (result == NULL)
|
302 |
|
|
return NULL;
|
303 |
|
|
result->entries = (PTR *) (*alloc_f) (size, sizeof (PTR));
|
304 |
|
|
if (result->entries == NULL)
|
305 |
|
|
{
|
306 |
|
|
if (free_f != NULL)
|
307 |
|
|
(*free_f) (result);
|
308 |
|
|
return NULL;
|
309 |
|
|
}
|
310 |
|
|
result->size = size;
|
311 |
|
|
result->size_prime_index = size_prime_index;
|
312 |
|
|
result->hash_f = hash_f;
|
313 |
|
|
result->eq_f = eq_f;
|
314 |
|
|
result->del_f = del_f;
|
315 |
|
|
result->alloc_f = alloc_f;
|
316 |
|
|
result->free_f = free_f;
|
317 |
|
|
return result;
|
318 |
|
|
}
|
319 |
|
|
|
320 |
|
|
/* As above, but use the variants of alloc_f and free_f which accept
|
321 |
|
|
an extra argument. */
|
322 |
|
|
|
323 |
|
|
htab_t
|
324 |
|
|
htab_create_alloc_ex (size_t size, htab_hash hash_f, htab_eq eq_f,
|
325 |
|
|
htab_del del_f, void *alloc_arg,
|
326 |
|
|
htab_alloc_with_arg alloc_f,
|
327 |
|
|
htab_free_with_arg free_f)
|
328 |
|
|
{
|
329 |
|
|
htab_t result;
|
330 |
|
|
unsigned int size_prime_index;
|
331 |
|
|
|
332 |
|
|
size_prime_index = higher_prime_index (size);
|
333 |
|
|
size = prime_tab[size_prime_index].prime;
|
334 |
|
|
|
335 |
|
|
result = (htab_t) (*alloc_f) (alloc_arg, 1, sizeof (struct htab));
|
336 |
|
|
if (result == NULL)
|
337 |
|
|
return NULL;
|
338 |
|
|
result->entries = (PTR *) (*alloc_f) (alloc_arg, size, sizeof (PTR));
|
339 |
|
|
if (result->entries == NULL)
|
340 |
|
|
{
|
341 |
|
|
if (free_f != NULL)
|
342 |
|
|
(*free_f) (alloc_arg, result);
|
343 |
|
|
return NULL;
|
344 |
|
|
}
|
345 |
|
|
result->size = size;
|
346 |
|
|
result->size_prime_index = size_prime_index;
|
347 |
|
|
result->hash_f = hash_f;
|
348 |
|
|
result->eq_f = eq_f;
|
349 |
|
|
result->del_f = del_f;
|
350 |
|
|
result->alloc_arg = alloc_arg;
|
351 |
|
|
result->alloc_with_arg_f = alloc_f;
|
352 |
|
|
result->free_with_arg_f = free_f;
|
353 |
|
|
return result;
|
354 |
|
|
}
|
355 |
|
|
|
356 |
|
|
/* Update the function pointers and allocation parameter in the htab_t. */
|
357 |
|
|
|
358 |
|
|
void
|
359 |
|
|
htab_set_functions_ex (htab_t htab, htab_hash hash_f, htab_eq eq_f,
|
360 |
|
|
htab_del del_f, PTR alloc_arg,
|
361 |
|
|
htab_alloc_with_arg alloc_f, htab_free_with_arg free_f)
|
362 |
|
|
{
|
363 |
|
|
htab->hash_f = hash_f;
|
364 |
|
|
htab->eq_f = eq_f;
|
365 |
|
|
htab->del_f = del_f;
|
366 |
|
|
htab->alloc_arg = alloc_arg;
|
367 |
|
|
htab->alloc_with_arg_f = alloc_f;
|
368 |
|
|
htab->free_with_arg_f = free_f;
|
369 |
|
|
}
|
370 |
|
|
|
371 |
|
|
/* These functions exist solely for backward compatibility. */
|
372 |
|
|
|
373 |
|
|
#undef htab_create
|
374 |
|
|
htab_t
|
375 |
|
|
htab_create (size_t size, htab_hash hash_f, htab_eq eq_f, htab_del del_f)
|
376 |
|
|
{
|
377 |
|
|
return htab_create_alloc (size, hash_f, eq_f, del_f, xcalloc, free);
|
378 |
|
|
}
|
379 |
|
|
|
380 |
|
|
htab_t
|
381 |
|
|
htab_try_create (size_t size, htab_hash hash_f, htab_eq eq_f, htab_del del_f)
|
382 |
|
|
{
|
383 |
|
|
return htab_create_alloc (size, hash_f, eq_f, del_f, calloc, free);
|
384 |
|
|
}
|
385 |
|
|
|
386 |
|
|
/* This function frees all memory allocated for given hash table.
|
387 |
|
|
Naturally the hash table must already exist. */
|
388 |
|
|
|
389 |
|
|
void
|
390 |
|
|
htab_delete (htab_t htab)
|
391 |
|
|
{
|
392 |
|
|
size_t size = htab_size (htab);
|
393 |
|
|
PTR *entries = htab->entries;
|
394 |
|
|
int i;
|
395 |
|
|
|
396 |
|
|
if (htab->del_f)
|
397 |
|
|
for (i = size - 1; i >= 0; i--)
|
398 |
|
|
if (entries[i] != HTAB_EMPTY_ENTRY && entries[i] != HTAB_DELETED_ENTRY)
|
399 |
|
|
(*htab->del_f) (entries[i]);
|
400 |
|
|
|
401 |
|
|
if (htab->free_f != NULL)
|
402 |
|
|
{
|
403 |
|
|
(*htab->free_f) (entries);
|
404 |
|
|
(*htab->free_f) (htab);
|
405 |
|
|
}
|
406 |
|
|
else if (htab->free_with_arg_f != NULL)
|
407 |
|
|
{
|
408 |
|
|
(*htab->free_with_arg_f) (htab->alloc_arg, entries);
|
409 |
|
|
(*htab->free_with_arg_f) (htab->alloc_arg, htab);
|
410 |
|
|
}
|
411 |
|
|
}
|
412 |
|
|
|
413 |
|
|
/* This function clears all entries in the given hash table. */
|
414 |
|
|
|
415 |
|
|
void
|
416 |
|
|
htab_empty (htab_t htab)
|
417 |
|
|
{
|
418 |
|
|
size_t size = htab_size (htab);
|
419 |
|
|
PTR *entries = htab->entries;
|
420 |
|
|
int i;
|
421 |
|
|
|
422 |
|
|
if (htab->del_f)
|
423 |
|
|
for (i = size - 1; i >= 0; i--)
|
424 |
|
|
if (entries[i] != HTAB_EMPTY_ENTRY && entries[i] != HTAB_DELETED_ENTRY)
|
425 |
|
|
(*htab->del_f) (entries[i]);
|
426 |
|
|
|
427 |
|
|
/* Instead of clearing megabyte, downsize the table. */
|
428 |
|
|
if (size > 1024*1024 / sizeof (PTR))
|
429 |
|
|
{
|
430 |
|
|
int nindex = higher_prime_index (1024 / sizeof (PTR));
|
431 |
|
|
int nsize = prime_tab[nindex].prime;
|
432 |
|
|
|
433 |
|
|
if (htab->free_f != NULL)
|
434 |
|
|
(*htab->free_f) (htab->entries);
|
435 |
|
|
else if (htab->free_with_arg_f != NULL)
|
436 |
|
|
(*htab->free_with_arg_f) (htab->alloc_arg, htab->entries);
|
437 |
|
|
if (htab->alloc_with_arg_f != NULL)
|
438 |
|
|
htab->entries = (PTR *) (*htab->alloc_with_arg_f) (htab->alloc_arg, nsize,
|
439 |
|
|
sizeof (PTR *));
|
440 |
|
|
else
|
441 |
|
|
htab->entries = (PTR *) (*htab->alloc_f) (nsize, sizeof (PTR *));
|
442 |
|
|
htab->size = nsize;
|
443 |
|
|
htab->size_prime_index = nindex;
|
444 |
|
|
}
|
445 |
|
|
else
|
446 |
|
|
memset (entries, 0, size * sizeof (PTR));
|
447 |
|
|
htab->n_deleted = 0;
|
448 |
|
|
htab->n_elements = 0;
|
449 |
|
|
}
|
450 |
|
|
|
451 |
|
|
/* Similar to htab_find_slot, but without several unwanted side effects:
|
452 |
|
|
- Does not call htab->eq_f when it finds an existing entry.
|
453 |
|
|
- Does not change the count of elements/searches/collisions in the
|
454 |
|
|
hash table.
|
455 |
|
|
This function also assumes there are no deleted entries in the table.
|
456 |
|
|
HASH is the hash value for the element to be inserted. */
|
457 |
|
|
|
458 |
|
|
static PTR *
|
459 |
|
|
find_empty_slot_for_expand (htab_t htab, hashval_t hash)
|
460 |
|
|
{
|
461 |
|
|
hashval_t index = htab_mod (hash, htab);
|
462 |
|
|
size_t size = htab_size (htab);
|
463 |
|
|
PTR *slot = htab->entries + index;
|
464 |
|
|
hashval_t hash2;
|
465 |
|
|
|
466 |
|
|
if (*slot == HTAB_EMPTY_ENTRY)
|
467 |
|
|
return slot;
|
468 |
|
|
else if (*slot == HTAB_DELETED_ENTRY)
|
469 |
|
|
abort ();
|
470 |
|
|
|
471 |
|
|
hash2 = htab_mod_m2 (hash, htab);
|
472 |
|
|
for (;;)
|
473 |
|
|
{
|
474 |
|
|
index += hash2;
|
475 |
|
|
if (index >= size)
|
476 |
|
|
index -= size;
|
477 |
|
|
|
478 |
|
|
slot = htab->entries + index;
|
479 |
|
|
if (*slot == HTAB_EMPTY_ENTRY)
|
480 |
|
|
return slot;
|
481 |
|
|
else if (*slot == HTAB_DELETED_ENTRY)
|
482 |
|
|
abort ();
|
483 |
|
|
}
|
484 |
|
|
}
|
485 |
|
|
|
486 |
|
|
/* The following function changes size of memory allocated for the
|
487 |
|
|
entries and repeatedly inserts the table elements. The occupancy
|
488 |
|
|
of the table after the call will be about 50%. Naturally the hash
|
489 |
|
|
table must already exist. Remember also that the place of the
|
490 |
|
|
table entries is changed. If memory allocation failures are allowed,
|
491 |
|
|
this function will return zero, indicating that the table could not be
|
492 |
|
|
expanded. If all goes well, it will return a non-zero value. */
|
493 |
|
|
|
494 |
|
|
static int
|
495 |
|
|
htab_expand (htab_t htab)
|
496 |
|
|
{
|
497 |
|
|
PTR *oentries;
|
498 |
|
|
PTR *olimit;
|
499 |
|
|
PTR *p;
|
500 |
|
|
PTR *nentries;
|
501 |
|
|
size_t nsize, osize, elts;
|
502 |
|
|
unsigned int oindex, nindex;
|
503 |
|
|
|
504 |
|
|
oentries = htab->entries;
|
505 |
|
|
oindex = htab->size_prime_index;
|
506 |
|
|
osize = htab->size;
|
507 |
|
|
olimit = oentries + osize;
|
508 |
|
|
elts = htab_elements (htab);
|
509 |
|
|
|
510 |
|
|
/* Resize only when table after removal of unused elements is either
|
511 |
|
|
too full or too empty. */
|
512 |
|
|
if (elts * 2 > osize || (elts * 8 < osize && osize > 32))
|
513 |
|
|
{
|
514 |
|
|
nindex = higher_prime_index (elts * 2);
|
515 |
|
|
nsize = prime_tab[nindex].prime;
|
516 |
|
|
}
|
517 |
|
|
else
|
518 |
|
|
{
|
519 |
|
|
nindex = oindex;
|
520 |
|
|
nsize = osize;
|
521 |
|
|
}
|
522 |
|
|
|
523 |
|
|
if (htab->alloc_with_arg_f != NULL)
|
524 |
|
|
nentries = (PTR *) (*htab->alloc_with_arg_f) (htab->alloc_arg, nsize,
|
525 |
|
|
sizeof (PTR *));
|
526 |
|
|
else
|
527 |
|
|
nentries = (PTR *) (*htab->alloc_f) (nsize, sizeof (PTR *));
|
528 |
|
|
if (nentries == NULL)
|
529 |
|
|
return 0;
|
530 |
|
|
htab->entries = nentries;
|
531 |
|
|
htab->size = nsize;
|
532 |
|
|
htab->size_prime_index = nindex;
|
533 |
|
|
htab->n_elements -= htab->n_deleted;
|
534 |
|
|
htab->n_deleted = 0;
|
535 |
|
|
|
536 |
|
|
p = oentries;
|
537 |
|
|
do
|
538 |
|
|
{
|
539 |
|
|
PTR x = *p;
|
540 |
|
|
|
541 |
|
|
if (x != HTAB_EMPTY_ENTRY && x != HTAB_DELETED_ENTRY)
|
542 |
|
|
{
|
543 |
|
|
PTR *q = find_empty_slot_for_expand (htab, (*htab->hash_f) (x));
|
544 |
|
|
|
545 |
|
|
*q = x;
|
546 |
|
|
}
|
547 |
|
|
|
548 |
|
|
p++;
|
549 |
|
|
}
|
550 |
|
|
while (p < olimit);
|
551 |
|
|
|
552 |
|
|
if (htab->free_f != NULL)
|
553 |
|
|
(*htab->free_f) (oentries);
|
554 |
|
|
else if (htab->free_with_arg_f != NULL)
|
555 |
|
|
(*htab->free_with_arg_f) (htab->alloc_arg, oentries);
|
556 |
|
|
return 1;
|
557 |
|
|
}
|
558 |
|
|
|
559 |
|
|
/* This function searches for a hash table entry equal to the given
|
560 |
|
|
element. It cannot be used to insert or delete an element. */
|
561 |
|
|
|
562 |
|
|
PTR
|
563 |
|
|
htab_find_with_hash (htab_t htab, const PTR element, hashval_t hash)
|
564 |
|
|
{
|
565 |
|
|
hashval_t index, hash2;
|
566 |
|
|
size_t size;
|
567 |
|
|
PTR entry;
|
568 |
|
|
|
569 |
|
|
htab->searches++;
|
570 |
|
|
size = htab_size (htab);
|
571 |
|
|
index = htab_mod (hash, htab);
|
572 |
|
|
|
573 |
|
|
entry = htab->entries[index];
|
574 |
|
|
if (entry == HTAB_EMPTY_ENTRY
|
575 |
|
|
|| (entry != HTAB_DELETED_ENTRY && (*htab->eq_f) (entry, element)))
|
576 |
|
|
return entry;
|
577 |
|
|
|
578 |
|
|
hash2 = htab_mod_m2 (hash, htab);
|
579 |
|
|
for (;;)
|
580 |
|
|
{
|
581 |
|
|
htab->collisions++;
|
582 |
|
|
index += hash2;
|
583 |
|
|
if (index >= size)
|
584 |
|
|
index -= size;
|
585 |
|
|
|
586 |
|
|
entry = htab->entries[index];
|
587 |
|
|
if (entry == HTAB_EMPTY_ENTRY
|
588 |
|
|
|| (entry != HTAB_DELETED_ENTRY && (*htab->eq_f) (entry, element)))
|
589 |
|
|
return entry;
|
590 |
|
|
}
|
591 |
|
|
}
|
592 |
|
|
|
593 |
|
|
/* Like htab_find_slot_with_hash, but compute the hash value from the
|
594 |
|
|
element. */
|
595 |
|
|
|
596 |
|
|
PTR
|
597 |
|
|
htab_find (htab_t htab, const PTR element)
|
598 |
|
|
{
|
599 |
|
|
return htab_find_with_hash (htab, element, (*htab->hash_f) (element));
|
600 |
|
|
}
|
601 |
|
|
|
602 |
|
|
/* This function searches for a hash table slot containing an entry
|
603 |
|
|
equal to the given element. To delete an entry, call this with
|
604 |
|
|
insert=NO_INSERT, then call htab_clear_slot on the slot returned
|
605 |
|
|
(possibly after doing some checks). To insert an entry, call this
|
606 |
|
|
with insert=INSERT, then write the value you want into the returned
|
607 |
|
|
slot. When inserting an entry, NULL may be returned if memory
|
608 |
|
|
allocation fails. */
|
609 |
|
|
|
610 |
|
|
PTR *
|
611 |
|
|
htab_find_slot_with_hash (htab_t htab, const PTR element,
|
612 |
|
|
hashval_t hash, enum insert_option insert)
|
613 |
|
|
{
|
614 |
|
|
PTR *first_deleted_slot;
|
615 |
|
|
hashval_t index, hash2;
|
616 |
|
|
size_t size;
|
617 |
|
|
PTR entry;
|
618 |
|
|
|
619 |
|
|
size = htab_size (htab);
|
620 |
|
|
if (insert == INSERT && size * 3 <= htab->n_elements * 4)
|
621 |
|
|
{
|
622 |
|
|
if (htab_expand (htab) == 0)
|
623 |
|
|
return NULL;
|
624 |
|
|
size = htab_size (htab);
|
625 |
|
|
}
|
626 |
|
|
|
627 |
|
|
index = htab_mod (hash, htab);
|
628 |
|
|
|
629 |
|
|
htab->searches++;
|
630 |
|
|
first_deleted_slot = NULL;
|
631 |
|
|
|
632 |
|
|
entry = htab->entries[index];
|
633 |
|
|
if (entry == HTAB_EMPTY_ENTRY)
|
634 |
|
|
goto empty_entry;
|
635 |
|
|
else if (entry == HTAB_DELETED_ENTRY)
|
636 |
|
|
first_deleted_slot = &htab->entries[index];
|
637 |
|
|
else if ((*htab->eq_f) (entry, element))
|
638 |
|
|
return &htab->entries[index];
|
639 |
|
|
|
640 |
|
|
hash2 = htab_mod_m2 (hash, htab);
|
641 |
|
|
for (;;)
|
642 |
|
|
{
|
643 |
|
|
htab->collisions++;
|
644 |
|
|
index += hash2;
|
645 |
|
|
if (index >= size)
|
646 |
|
|
index -= size;
|
647 |
|
|
|
648 |
|
|
entry = htab->entries[index];
|
649 |
|
|
if (entry == HTAB_EMPTY_ENTRY)
|
650 |
|
|
goto empty_entry;
|
651 |
|
|
else if (entry == HTAB_DELETED_ENTRY)
|
652 |
|
|
{
|
653 |
|
|
if (!first_deleted_slot)
|
654 |
|
|
first_deleted_slot = &htab->entries[index];
|
655 |
|
|
}
|
656 |
|
|
else if ((*htab->eq_f) (entry, element))
|
657 |
|
|
return &htab->entries[index];
|
658 |
|
|
}
|
659 |
|
|
|
660 |
|
|
empty_entry:
|
661 |
|
|
if (insert == NO_INSERT)
|
662 |
|
|
return NULL;
|
663 |
|
|
|
664 |
|
|
if (first_deleted_slot)
|
665 |
|
|
{
|
666 |
|
|
htab->n_deleted--;
|
667 |
|
|
*first_deleted_slot = HTAB_EMPTY_ENTRY;
|
668 |
|
|
return first_deleted_slot;
|
669 |
|
|
}
|
670 |
|
|
|
671 |
|
|
htab->n_elements++;
|
672 |
|
|
return &htab->entries[index];
|
673 |
|
|
}
|
674 |
|
|
|
675 |
|
|
/* Like htab_find_slot_with_hash, but compute the hash value from the
|
676 |
|
|
element. */
|
677 |
|
|
|
678 |
|
|
PTR *
|
679 |
|
|
htab_find_slot (htab_t htab, const PTR element, enum insert_option insert)
|
680 |
|
|
{
|
681 |
|
|
return htab_find_slot_with_hash (htab, element, (*htab->hash_f) (element),
|
682 |
|
|
insert);
|
683 |
|
|
}
|
684 |
|
|
|
685 |
|
|
/* This function deletes an element with the given value from hash
|
686 |
|
|
table (the hash is computed from the element). If there is no matching
|
687 |
|
|
element in the hash table, this function does nothing. */
|
688 |
|
|
|
689 |
|
|
void
|
690 |
|
|
htab_remove_elt (htab_t htab, PTR element)
|
691 |
|
|
{
|
692 |
|
|
htab_remove_elt_with_hash (htab, element, (*htab->hash_f) (element));
|
693 |
|
|
}
|
694 |
|
|
|
695 |
|
|
|
696 |
|
|
/* This function deletes an element with the given value from hash
|
697 |
|
|
table. If there is no matching element in the hash table, this
|
698 |
|
|
function does nothing. */
|
699 |
|
|
|
700 |
|
|
void
|
701 |
|
|
htab_remove_elt_with_hash (htab_t htab, PTR element, hashval_t hash)
|
702 |
|
|
{
|
703 |
|
|
PTR *slot;
|
704 |
|
|
|
705 |
|
|
slot = htab_find_slot_with_hash (htab, element, hash, NO_INSERT);
|
706 |
|
|
if (*slot == HTAB_EMPTY_ENTRY)
|
707 |
|
|
return;
|
708 |
|
|
|
709 |
|
|
if (htab->del_f)
|
710 |
|
|
(*htab->del_f) (*slot);
|
711 |
|
|
|
712 |
|
|
*slot = HTAB_DELETED_ENTRY;
|
713 |
|
|
htab->n_deleted++;
|
714 |
|
|
}
|
715 |
|
|
|
716 |
|
|
/* This function clears a specified slot in a hash table. It is
|
717 |
|
|
useful when you've already done the lookup and don't want to do it
|
718 |
|
|
again. */
|
719 |
|
|
|
720 |
|
|
void
|
721 |
|
|
htab_clear_slot (htab_t htab, PTR *slot)
|
722 |
|
|
{
|
723 |
|
|
if (slot < htab->entries || slot >= htab->entries + htab_size (htab)
|
724 |
|
|
|| *slot == HTAB_EMPTY_ENTRY || *slot == HTAB_DELETED_ENTRY)
|
725 |
|
|
abort ();
|
726 |
|
|
|
727 |
|
|
if (htab->del_f)
|
728 |
|
|
(*htab->del_f) (*slot);
|
729 |
|
|
|
730 |
|
|
*slot = HTAB_DELETED_ENTRY;
|
731 |
|
|
htab->n_deleted++;
|
732 |
|
|
}
|
733 |
|
|
|
734 |
|
|
/* This function scans over the entire hash table calling
|
735 |
|
|
CALLBACK for each live entry. If CALLBACK returns false,
|
736 |
|
|
the iteration stops. INFO is passed as CALLBACK's second
|
737 |
|
|
argument. */
|
738 |
|
|
|
739 |
|
|
void
|
740 |
|
|
htab_traverse_noresize (htab_t htab, htab_trav callback, PTR info)
|
741 |
|
|
{
|
742 |
|
|
PTR *slot;
|
743 |
|
|
PTR *limit;
|
744 |
|
|
|
745 |
|
|
slot = htab->entries;
|
746 |
|
|
limit = slot + htab_size (htab);
|
747 |
|
|
|
748 |
|
|
do
|
749 |
|
|
{
|
750 |
|
|
PTR x = *slot;
|
751 |
|
|
|
752 |
|
|
if (x != HTAB_EMPTY_ENTRY && x != HTAB_DELETED_ENTRY)
|
753 |
|
|
if (!(*callback) (slot, info))
|
754 |
|
|
break;
|
755 |
|
|
}
|
756 |
|
|
while (++slot < limit);
|
757 |
|
|
}
|
758 |
|
|
|
759 |
|
|
/* Like htab_traverse_noresize, but does resize the table when it is
|
760 |
|
|
too empty to improve effectivity of subsequent calls. */
|
761 |
|
|
|
762 |
|
|
void
|
763 |
|
|
htab_traverse (htab_t htab, htab_trav callback, PTR info)
|
764 |
|
|
{
|
765 |
|
|
size_t size = htab_size (htab);
|
766 |
|
|
if (htab_elements (htab) * 8 < size && size > 32)
|
767 |
|
|
htab_expand (htab);
|
768 |
|
|
|
769 |
|
|
htab_traverse_noresize (htab, callback, info);
|
770 |
|
|
}
|
771 |
|
|
|
772 |
|
|
/* Return the fraction of fixed collisions during all work with given
|
773 |
|
|
hash table. */
|
774 |
|
|
|
775 |
|
|
double
|
776 |
|
|
htab_collisions (htab_t htab)
|
777 |
|
|
{
|
778 |
|
|
if (htab->searches == 0)
|
779 |
|
|
return 0.0;
|
780 |
|
|
|
781 |
|
|
return (double) htab->collisions / (double) htab->searches;
|
782 |
|
|
}
|
783 |
|
|
|
784 |
|
|
/* Hash P as a null-terminated string.
|
785 |
|
|
|
786 |
|
|
Copied from gcc/hashtable.c. Zack had the following to say with respect
|
787 |
|
|
to applicability, though note that unlike hashtable.c, this hash table
|
788 |
|
|
implementation re-hashes rather than chain buckets.
|
789 |
|
|
|
790 |
|
|
http://gcc.gnu.org/ml/gcc-patches/2001-08/msg01021.html
|
791 |
|
|
From: Zack Weinberg <zackw@panix.com>
|
792 |
|
|
Date: Fri, 17 Aug 2001 02:15:56 -0400
|
793 |
|
|
|
794 |
|
|
I got it by extracting all the identifiers from all the source code
|
795 |
|
|
I had lying around in mid-1999, and testing many recurrences of
|
796 |
|
|
the form "H_n = H_{n-1} * K + c_n * L + M" where K, L, M were either
|
797 |
|
|
prime numbers or the appropriate identity. This was the best one.
|
798 |
|
|
I don't remember exactly what constituted "best", except I was
|
799 |
|
|
looking at bucket-length distributions mostly.
|
800 |
|
|
|
801 |
|
|
So it should be very good at hashing identifiers, but might not be
|
802 |
|
|
as good at arbitrary strings.
|
803 |
|
|
|
804 |
|
|
I'll add that it thoroughly trounces the hash functions recommended
|
805 |
|
|
for this use at http://burtleburtle.net/bob/hash/index.html, both
|
806 |
|
|
on speed and bucket distribution. I haven't tried it against the
|
807 |
|
|
function they just started using for Perl's hashes. */
|
808 |
|
|
|
809 |
|
|
hashval_t
|
810 |
|
|
htab_hash_string (const PTR p)
|
811 |
|
|
{
|
812 |
|
|
const unsigned char *str = (const unsigned char *) p;
|
813 |
|
|
hashval_t r = 0;
|
814 |
|
|
unsigned char c;
|
815 |
|
|
|
816 |
|
|
while ((c = *str++) != 0)
|
817 |
|
|
r = r * 67 + c - 113;
|
818 |
|
|
|
819 |
|
|
return r;
|
820 |
|
|
}
|
821 |
|
|
|
822 |
|
|
/* DERIVED FROM:
|
823 |
|
|
--------------------------------------------------------------------
|
824 |
|
|
lookup2.c, by Bob Jenkins, December 1996, Public Domain.
|
825 |
|
|
hash(), hash2(), hash3, and mix() are externally useful functions.
|
826 |
|
|
Routines to test the hash are included if SELF_TEST is defined.
|
827 |
|
|
You can use this free for any purpose. It has no warranty.
|
828 |
|
|
--------------------------------------------------------------------
|
829 |
|
|
*/
|
830 |
|
|
|
831 |
|
|
/*
|
832 |
|
|
--------------------------------------------------------------------
|
833 |
|
|
mix -- mix 3 32-bit values reversibly.
|
834 |
|
|
For every delta with one or two bit set, and the deltas of all three
|
835 |
|
|
high bits or all three low bits, whether the original value of a,b,c
|
836 |
|
|
is almost all zero or is uniformly distributed,
|
837 |
|
|
* If mix() is run forward or backward, at least 32 bits in a,b,c
|
838 |
|
|
have at least 1/4 probability of changing.
|
839 |
|
|
* If mix() is run forward, every bit of c will change between 1/3 and
|
840 |
|
|
2/3 of the time. (Well, 22/100 and 78/100 for some 2-bit deltas.)
|
841 |
|
|
mix() was built out of 36 single-cycle latency instructions in a
|
842 |
|
|
structure that could supported 2x parallelism, like so:
|
843 |
|
|
a -= b;
|
844 |
|
|
a -= c; x = (c>>13);
|
845 |
|
|
b -= c; a ^= x;
|
846 |
|
|
b -= a; x = (a<<8);
|
847 |
|
|
c -= a; b ^= x;
|
848 |
|
|
c -= b; x = (b>>13);
|
849 |
|
|
...
|
850 |
|
|
Unfortunately, superscalar Pentiums and Sparcs can't take advantage
|
851 |
|
|
of that parallelism. They've also turned some of those single-cycle
|
852 |
|
|
latency instructions into multi-cycle latency instructions. Still,
|
853 |
|
|
this is the fastest good hash I could find. There were about 2^^68
|
854 |
|
|
to choose from. I only looked at a billion or so.
|
855 |
|
|
--------------------------------------------------------------------
|
856 |
|
|
*/
|
857 |
|
|
/* same, but slower, works on systems that might have 8 byte hashval_t's */
|
858 |
|
|
#define mix(a,b,c) \
|
859 |
|
|
{ \
|
860 |
|
|
a -= b; a -= c; a ^= (c>>13); \
|
861 |
|
|
b -= c; b -= a; b ^= (a<< 8); \
|
862 |
|
|
c -= a; c -= b; c ^= ((b&0xffffffff)>>13); \
|
863 |
|
|
a -= b; a -= c; a ^= ((c&0xffffffff)>>12); \
|
864 |
|
|
b -= c; b -= a; b = (b ^ (a<<16)) & 0xffffffff; \
|
865 |
|
|
c -= a; c -= b; c = (c ^ (b>> 5)) & 0xffffffff; \
|
866 |
|
|
a -= b; a -= c; a = (a ^ (c>> 3)) & 0xffffffff; \
|
867 |
|
|
b -= c; b -= a; b = (b ^ (a<<10)) & 0xffffffff; \
|
868 |
|
|
c -= a; c -= b; c = (c ^ (b>>15)) & 0xffffffff; \
|
869 |
|
|
}
|
870 |
|
|
|
871 |
|
|
/*
|
872 |
|
|
--------------------------------------------------------------------
|
873 |
|
|
hash() -- hash a variable-length key into a 32-bit value
|
874 |
|
|
k : the key (the unaligned variable-length array of bytes)
|
875 |
|
|
len : the length of the key, counting by bytes
|
876 |
|
|
level : can be any 4-byte value
|
877 |
|
|
Returns a 32-bit value. Every bit of the key affects every bit of
|
878 |
|
|
the return value. Every 1-bit and 2-bit delta achieves avalanche.
|
879 |
|
|
About 36+6len instructions.
|
880 |
|
|
|
881 |
|
|
The best hash table sizes are powers of 2. There is no need to do
|
882 |
|
|
mod a prime (mod is sooo slow!). If you need less than 32 bits,
|
883 |
|
|
use a bitmask. For example, if you need only 10 bits, do
|
884 |
|
|
h = (h & hashmask(10));
|
885 |
|
|
In which case, the hash table should have hashsize(10) elements.
|
886 |
|
|
|
887 |
|
|
If you are hashing n strings (ub1 **)k, do it like this:
|
888 |
|
|
for (i=0, h=0; i<n; ++i) h = hash( k[i], len[i], h);
|
889 |
|
|
|
890 |
|
|
By Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may use this
|
891 |
|
|
code any way you wish, private, educational, or commercial. It's free.
|
892 |
|
|
|
893 |
|
|
See http://burtleburtle.net/bob/hash/evahash.html
|
894 |
|
|
Use for hash table lookup, or anything where one collision in 2^32 is
|
895 |
|
|
acceptable. Do NOT use for cryptographic purposes.
|
896 |
|
|
--------------------------------------------------------------------
|
897 |
|
|
*/
|
898 |
|
|
|
899 |
|
|
hashval_t
|
900 |
|
|
iterative_hash (const PTR k_in /* the key */,
|
901 |
|
|
register size_t length /* the length of the key */,
|
902 |
|
|
register hashval_t initval /* the previous hash, or
|
903 |
|
|
an arbitrary value */)
|
904 |
|
|
{
|
905 |
|
|
register const unsigned char *k = (const unsigned char *)k_in;
|
906 |
|
|
register hashval_t a,b,c,len;
|
907 |
|
|
|
908 |
|
|
/* Set up the internal state */
|
909 |
|
|
len = length;
|
910 |
|
|
a = b = 0x9e3779b9; /* the golden ratio; an arbitrary value */
|
911 |
|
|
c = initval; /* the previous hash value */
|
912 |
|
|
|
913 |
|
|
/*---------------------------------------- handle most of the key */
|
914 |
|
|
#ifndef WORDS_BIGENDIAN
|
915 |
|
|
/* On a little-endian machine, if the data is 4-byte aligned we can hash
|
916 |
|
|
by word for better speed. This gives nondeterministic results on
|
917 |
|
|
big-endian machines. */
|
918 |
|
|
if (sizeof (hashval_t) == 4 && (((size_t)k)&3) == 0)
|
919 |
|
|
while (len >= 12) /* aligned */
|
920 |
|
|
{
|
921 |
|
|
a += *(hashval_t *)(k+0);
|
922 |
|
|
b += *(hashval_t *)(k+4);
|
923 |
|
|
c += *(hashval_t *)(k+8);
|
924 |
|
|
mix(a,b,c);
|
925 |
|
|
k += 12; len -= 12;
|
926 |
|
|
}
|
927 |
|
|
else /* unaligned */
|
928 |
|
|
#endif
|
929 |
|
|
while (len >= 12)
|
930 |
|
|
{
|
931 |
|
|
a += (k[0] +((hashval_t)k[1]<<8) +((hashval_t)k[2]<<16) +((hashval_t)k[3]<<24));
|
932 |
|
|
b += (k[4] +((hashval_t)k[5]<<8) +((hashval_t)k[6]<<16) +((hashval_t)k[7]<<24));
|
933 |
|
|
c += (k[8] +((hashval_t)k[9]<<8) +((hashval_t)k[10]<<16)+((hashval_t)k[11]<<24));
|
934 |
|
|
mix(a,b,c);
|
935 |
|
|
k += 12; len -= 12;
|
936 |
|
|
}
|
937 |
|
|
|
938 |
|
|
/*------------------------------------- handle the last 11 bytes */
|
939 |
|
|
c += length;
|
940 |
|
|
switch(len) /* all the case statements fall through */
|
941 |
|
|
{
|
942 |
|
|
case 11: c+=((hashval_t)k[10]<<24);
|
943 |
|
|
case 10: c+=((hashval_t)k[9]<<16);
|
944 |
|
|
case 9 : c+=((hashval_t)k[8]<<8);
|
945 |
|
|
/* the first byte of c is reserved for the length */
|
946 |
|
|
case 8 : b+=((hashval_t)k[7]<<24);
|
947 |
|
|
case 7 : b+=((hashval_t)k[6]<<16);
|
948 |
|
|
case 6 : b+=((hashval_t)k[5]<<8);
|
949 |
|
|
case 5 : b+=k[4];
|
950 |
|
|
case 4 : a+=((hashval_t)k[3]<<24);
|
951 |
|
|
case 3 : a+=((hashval_t)k[2]<<16);
|
952 |
|
|
case 2 : a+=((hashval_t)k[1]<<8);
|
953 |
|
|
case 1 : a+=k[0];
|
954 |
|
|
/* case 0: nothing left to add */
|
955 |
|
|
}
|
956 |
|
|
mix(a,b,c);
|
957 |
|
|
/*-------------------------------------------- report the result */
|
958 |
|
|
return c;
|
959 |
|
|
}
|