]>
Commit | Line | Data |
---|---|---|
a2f945c6 | 1 | /* An expandable hash tables datatype. |
9bf3c9cc RH |
2 | Copyright (C) 1999, 2000, 2001, 2002, 2003, 2004 |
3 | Free Software Foundation, Inc. | |
a2f945c6 VM |
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 | |
ee58dffd NC |
19 | not, write to the Free Software Foundation, Inc., 51 Franklin Street - Fifth Floor, |
20 | Boston, MA 02110-1301, USA. */ | |
a2f945c6 VM |
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 | ||
6de9b8ff PDM |
39 | #include <sys/types.h> |
40 | ||
a2f945c6 VM |
41 | #ifdef HAVE_STDLIB_H |
42 | #include <stdlib.h> | |
43 | #endif | |
d11ec6f0 ZW |
44 | #ifdef HAVE_STRING_H |
45 | #include <string.h> | |
46 | #endif | |
cf8e4b78 DH |
47 | #ifdef HAVE_MALLOC_H |
48 | #include <malloc.h> | |
49 | #endif | |
9bf3c9cc RH |
50 | #ifdef HAVE_LIMITS_H |
51 | #include <limits.h> | |
52 | #endif | |
53 | #ifdef HAVE_STDINT_H | |
54 | #include <stdint.h> | |
55 | #endif | |
cf8e4b78 | 56 | |
36dd3a44 JL |
57 | #include <stdio.h> |
58 | ||
a2f945c6 | 59 | #include "libiberty.h" |
9bf3c9cc | 60 | #include "ansidecl.h" |
a2f945c6 VM |
61 | #include "hashtab.h" |
62 | ||
9bf3c9cc RH |
63 | #ifndef CHAR_BIT |
64 | #define CHAR_BIT 8 | |
65 | #endif | |
66 | ||
a2f945c6 VM |
67 | /* This macro defines reserved value for empty table entry. */ |
68 | ||
35e9340f | 69 | #define EMPTY_ENTRY ((PTR) 0) |
a2f945c6 VM |
70 | |
71 | /* This macro defines reserved value for table entry which contained | |
72 | a deleted element. */ | |
73 | ||
35e9340f | 74 | #define DELETED_ENTRY ((PTR) 1) |
a2f945c6 | 75 | |
6da879de GDR |
76 | static unsigned int higher_prime_index (unsigned long); |
77 | static hashval_t htab_mod_1 (hashval_t, hashval_t, hashval_t, int); | |
78 | static hashval_t htab_mod (hashval_t, htab_t); | |
79 | static hashval_t htab_mod_m2 (hashval_t, htab_t); | |
80 | static hashval_t hash_pointer (const void *); | |
81 | static int eq_pointer (const void *, const void *); | |
82 | static int htab_expand (htab_t); | |
83 | static PTR *find_empty_slot_for_expand (htab_t, hashval_t); | |
18a94a2f MM |
84 | |
85 | /* At some point, we could make these be NULL, and modify the | |
86 | hash-table routines to handle NULL specially; that would avoid | |
87 | function-call overhead for the common case of hashing pointers. */ | |
88 | htab_hash htab_hash_pointer = hash_pointer; | |
89 | htab_eq htab_eq_pointer = eq_pointer; | |
0194e877 | 90 | |
9bf3c9cc RH |
91 | /* Table of primes and multiplicative inverses. |
92 | ||
93 | Note that these are not minimally reduced inverses. Unlike when generating | |
94 | code to divide by a constant, we want to be able to use the same algorithm | |
95 | all the time. All of these inverses (are implied to) have bit 32 set. | |
96 | ||
97 | For the record, here's the function that computed the table; it's a | |
98 | vastly simplified version of the function of the same name from gcc. */ | |
99 | ||
100 | #if 0 | |
101 | unsigned int | |
102 | ceil_log2 (unsigned int x) | |
103 | { | |
104 | int i; | |
105 | for (i = 31; i >= 0 ; --i) | |
106 | if (x > (1u << i)) | |
107 | return i+1; | |
108 | abort (); | |
109 | } | |
a2f945c6 | 110 | |
9bf3c9cc RH |
111 | unsigned int |
112 | choose_multiplier (unsigned int d, unsigned int *mlp, unsigned char *shiftp) | |
113 | { | |
114 | unsigned long long mhigh; | |
115 | double nx; | |
116 | int lgup, post_shift; | |
117 | int pow, pow2; | |
118 | int n = 32, precision = 32; | |
119 | ||
120 | lgup = ceil_log2 (d); | |
121 | pow = n + lgup; | |
122 | pow2 = n + lgup - precision; | |
123 | ||
124 | nx = ldexp (1.0, pow) + ldexp (1.0, pow2); | |
125 | mhigh = nx / d; | |
126 | ||
127 | *shiftp = lgup - 1; | |
128 | *mlp = mhigh; | |
129 | return mhigh >> 32; | |
130 | } | |
131 | #endif | |
132 | ||
133 | struct prime_ent | |
134 | { | |
135 | hashval_t prime; | |
136 | hashval_t inv; | |
137 | hashval_t inv_m2; /* inverse of prime-2 */ | |
138 | hashval_t shift; | |
139 | }; | |
140 | ||
141 | static struct prime_ent const prime_tab[] = { | |
142 | { 7, 0x24924925, 0x9999999b, 2 }, | |
143 | { 13, 0x3b13b13c, 0x745d1747, 3 }, | |
144 | { 31, 0x08421085, 0x1a7b9612, 4 }, | |
145 | { 61, 0x0c9714fc, 0x15b1e5f8, 5 }, | |
146 | { 127, 0x02040811, 0x0624dd30, 6 }, | |
147 | { 251, 0x05197f7e, 0x073260a5, 7 }, | |
148 | { 509, 0x01824366, 0x02864fc8, 8 }, | |
149 | { 1021, 0x00c0906d, 0x014191f7, 9 }, | |
150 | { 2039, 0x0121456f, 0x0161e69e, 10 }, | |
151 | { 4093, 0x00300902, 0x00501908, 11 }, | |
152 | { 8191, 0x00080041, 0x00180241, 12 }, | |
153 | { 16381, 0x000c0091, 0x00140191, 13 }, | |
154 | { 32749, 0x002605a5, 0x002a06e6, 14 }, | |
155 | { 65521, 0x000f00e2, 0x00110122, 15 }, | |
156 | { 131071, 0x00008001, 0x00018003, 16 }, | |
157 | { 262139, 0x00014002, 0x0001c004, 17 }, | |
158 | { 524287, 0x00002001, 0x00006001, 18 }, | |
159 | { 1048573, 0x00003001, 0x00005001, 19 }, | |
160 | { 2097143, 0x00004801, 0x00005801, 20 }, | |
161 | { 4194301, 0x00000c01, 0x00001401, 21 }, | |
162 | { 8388593, 0x00001e01, 0x00002201, 22 }, | |
163 | { 16777213, 0x00000301, 0x00000501, 23 }, | |
164 | { 33554393, 0x00001381, 0x00001481, 24 }, | |
165 | { 67108859, 0x00000141, 0x000001c1, 25 }, | |
166 | { 134217689, 0x000004e1, 0x00000521, 26 }, | |
167 | { 268435399, 0x00000391, 0x000003b1, 27 }, | |
168 | { 536870909, 0x00000019, 0x00000029, 28 }, | |
169 | { 1073741789, 0x0000008d, 0x00000095, 29 }, | |
170 | { 2147483647, 0x00000003, 0x00000007, 30 }, | |
171 | /* Avoid "decimal constant so large it is unsigned" for 4294967291. */ | |
172 | { 0xfffffffb, 0x00000006, 0x00000008, 31 } | |
173 | }; | |
174 | ||
175 | /* The following function returns an index into the above table of the | |
176 | nearest prime number which is greater than N, and near a power of two. */ | |
177 | ||
178 | static unsigned int | |
6da879de | 179 | higher_prime_index (unsigned long n) |
a2f945c6 | 180 | { |
9bf3c9cc RH |
181 | unsigned int low = 0; |
182 | unsigned int high = sizeof(prime_tab) / sizeof(prime_tab[0]); | |
a4c9b97e MM |
183 | |
184 | while (low != high) | |
185 | { | |
9bf3c9cc RH |
186 | unsigned int mid = low + (high - low) / 2; |
187 | if (n > prime_tab[mid].prime) | |
a4c9b97e MM |
188 | low = mid + 1; |
189 | else | |
190 | high = mid; | |
191 | } | |
192 | ||
193 | /* If we've run out of primes, abort. */ | |
9bf3c9cc | 194 | if (n > prime_tab[low].prime) |
a4c9b97e MM |
195 | { |
196 | fprintf (stderr, "Cannot find prime bigger than %lu\n", n); | |
197 | abort (); | |
198 | } | |
199 | ||
9bf3c9cc | 200 | return low; |
a2f945c6 VM |
201 | } |
202 | ||
18a94a2f MM |
203 | /* Returns a hash code for P. */ |
204 | ||
4feeaae3 | 205 | static hashval_t |
6da879de | 206 | hash_pointer (const PTR p) |
18a94a2f | 207 | { |
1d2da2e1 | 208 | return (hashval_t) ((long)p >> 3); |
18a94a2f MM |
209 | } |
210 | ||
211 | /* Returns non-zero if P1 and P2 are equal. */ | |
212 | ||
4feeaae3 | 213 | static int |
6da879de | 214 | eq_pointer (const PTR p1, const PTR p2) |
18a94a2f MM |
215 | { |
216 | return p1 == p2; | |
217 | } | |
218 | ||
d9175b87 | 219 | |
d7cf8390 GDR |
220 | /* The parens around the function names in the next two definitions |
221 | are essential in order to prevent macro expansions of the name. | |
222 | The bodies, however, are expanded as expected, so they are not | |
223 | recursive definitions. */ | |
224 | ||
225 | /* Return the current size of given hash table. */ | |
226 | ||
227 | #define htab_size(htab) ((htab)->size) | |
228 | ||
229 | size_t | |
230 | (htab_size) (htab_t htab) | |
d9175b87 | 231 | { |
d7cf8390 | 232 | return htab_size (htab); |
d9175b87 RH |
233 | } |
234 | ||
235 | /* Return the current number of elements in given hash table. */ | |
236 | ||
d7cf8390 GDR |
237 | #define htab_elements(htab) ((htab)->n_elements - (htab)->n_deleted) |
238 | ||
239 | size_t | |
240 | (htab_elements) (htab_t htab) | |
d9175b87 | 241 | { |
d7cf8390 | 242 | return htab_elements (htab); |
d9175b87 RH |
243 | } |
244 | ||
9bf3c9cc RH |
245 | /* Return X % Y. */ |
246 | ||
247 | static inline hashval_t | |
6da879de | 248 | htab_mod_1 (hashval_t x, hashval_t y, hashval_t inv, int shift) |
9bf3c9cc RH |
249 | { |
250 | /* The multiplicative inverses computed above are for 32-bit types, and | |
251 | requires that we be able to compute a highpart multiply. */ | |
252 | #ifdef UNSIGNED_64BIT_TYPE | |
253 | __extension__ typedef UNSIGNED_64BIT_TYPE ull; | |
254 | if (sizeof (hashval_t) * CHAR_BIT <= 32) | |
255 | { | |
256 | hashval_t t1, t2, t3, t4, q, r; | |
257 | ||
258 | t1 = ((ull)x * inv) >> 32; | |
259 | t2 = x - t1; | |
260 | t3 = t2 >> 1; | |
261 | t4 = t1 + t3; | |
262 | q = t4 >> shift; | |
263 | r = x - (q * y); | |
264 | ||
265 | return r; | |
266 | } | |
267 | #endif | |
268 | ||
269 | /* Otherwise just use the native division routines. */ | |
270 | return x % y; | |
271 | } | |
272 | ||
d9175b87 RH |
273 | /* Compute the primary hash for HASH given HTAB's current size. */ |
274 | ||
275 | static inline hashval_t | |
6da879de | 276 | htab_mod (hashval_t hash, htab_t htab) |
d9175b87 | 277 | { |
9bf3c9cc RH |
278 | const struct prime_ent *p = &prime_tab[htab->size_prime_index]; |
279 | return htab_mod_1 (hash, p->prime, p->inv, p->shift); | |
d9175b87 RH |
280 | } |
281 | ||
282 | /* Compute the secondary hash for HASH given HTAB's current size. */ | |
283 | ||
284 | static inline hashval_t | |
6da879de | 285 | htab_mod_m2 (hashval_t hash, htab_t htab) |
d9175b87 | 286 | { |
9bf3c9cc RH |
287 | const struct prime_ent *p = &prime_tab[htab->size_prime_index]; |
288 | return 1 + htab_mod_1 (hash, p->prime - 2, p->inv_m2, p->shift); | |
d9175b87 RH |
289 | } |
290 | ||
a2f945c6 VM |
291 | /* This function creates table with length slightly longer than given |
292 | source length. Created hash table is initiated as empty (all the | |
293 | hash table entries are EMPTY_ENTRY). The function returns the | |
e2500fed | 294 | created hash table, or NULL if memory allocation fails. */ |
a2f945c6 | 295 | |
5194cf08 | 296 | htab_t |
6da879de GDR |
297 | htab_create_alloc (size_t size, htab_hash hash_f, htab_eq eq_f, |
298 | htab_del del_f, htab_alloc alloc_f, htab_free free_f) | |
a2f945c6 | 299 | { |
5194cf08 | 300 | htab_t result; |
9bf3c9cc RH |
301 | unsigned int size_prime_index; |
302 | ||
303 | size_prime_index = higher_prime_index (size); | |
304 | size = prime_tab[size_prime_index].prime; | |
a2f945c6 | 305 | |
e2500fed | 306 | result = (htab_t) (*alloc_f) (1, sizeof (struct htab)); |
d50d20ec HPN |
307 | if (result == NULL) |
308 | return NULL; | |
e2500fed | 309 | result->entries = (PTR *) (*alloc_f) (size, sizeof (PTR)); |
d50d20ec HPN |
310 | if (result->entries == NULL) |
311 | { | |
e2500fed GK |
312 | if (free_f != NULL) |
313 | (*free_f) (result); | |
d50d20ec HPN |
314 | return NULL; |
315 | } | |
d50d20ec | 316 | result->size = size; |
9bf3c9cc | 317 | result->size_prime_index = size_prime_index; |
d50d20ec HPN |
318 | result->hash_f = hash_f; |
319 | result->eq_f = eq_f; | |
320 | result->del_f = del_f; | |
e2500fed GK |
321 | result->alloc_f = alloc_f; |
322 | result->free_f = free_f; | |
a2f945c6 VM |
323 | return result; |
324 | } | |
325 | ||
74828682 DJ |
326 | /* As above, but use the variants of alloc_f and free_f which accept |
327 | an extra argument. */ | |
328 | ||
329 | htab_t | |
d7cf8390 GDR |
330 | htab_create_alloc_ex (size_t size, htab_hash hash_f, htab_eq eq_f, |
331 | htab_del del_f, void *alloc_arg, | |
332 | htab_alloc_with_arg alloc_f, | |
333 | htab_free_with_arg free_f) | |
74828682 DJ |
334 | { |
335 | htab_t result; | |
9bf3c9cc RH |
336 | unsigned int size_prime_index; |
337 | ||
338 | size_prime_index = higher_prime_index (size); | |
339 | size = prime_tab[size_prime_index].prime; | |
74828682 | 340 | |
74828682 DJ |
341 | result = (htab_t) (*alloc_f) (alloc_arg, 1, sizeof (struct htab)); |
342 | if (result == NULL) | |
343 | return NULL; | |
344 | result->entries = (PTR *) (*alloc_f) (alloc_arg, size, sizeof (PTR)); | |
345 | if (result->entries == NULL) | |
346 | { | |
347 | if (free_f != NULL) | |
348 | (*free_f) (alloc_arg, result); | |
349 | return NULL; | |
350 | } | |
351 | result->size = size; | |
9bf3c9cc | 352 | result->size_prime_index = size_prime_index; |
74828682 DJ |
353 | result->hash_f = hash_f; |
354 | result->eq_f = eq_f; | |
355 | result->del_f = del_f; | |
356 | result->alloc_arg = alloc_arg; | |
357 | result->alloc_with_arg_f = alloc_f; | |
358 | result->free_with_arg_f = free_f; | |
359 | return result; | |
360 | } | |
361 | ||
362 | /* Update the function pointers and allocation parameter in the htab_t. */ | |
363 | ||
364 | void | |
6da879de GDR |
365 | htab_set_functions_ex (htab_t htab, htab_hash hash_f, htab_eq eq_f, |
366 | htab_del del_f, PTR alloc_arg, | |
367 | htab_alloc_with_arg alloc_f, htab_free_with_arg free_f) | |
74828682 DJ |
368 | { |
369 | htab->hash_f = hash_f; | |
370 | htab->eq_f = eq_f; | |
371 | htab->del_f = del_f; | |
372 | htab->alloc_arg = alloc_arg; | |
373 | htab->alloc_with_arg_f = alloc_f; | |
374 | htab->free_with_arg_f = free_f; | |
375 | } | |
376 | ||
045b3a49 GK |
377 | /* These functions exist solely for backward compatibility. */ |
378 | ||
379 | #undef htab_create | |
380 | htab_t | |
6da879de | 381 | htab_create (size_t size, htab_hash hash_f, htab_eq eq_f, htab_del del_f) |
045b3a49 GK |
382 | { |
383 | return htab_create_alloc (size, hash_f, eq_f, del_f, xcalloc, free); | |
384 | } | |
385 | ||
386 | htab_t | |
6da879de | 387 | htab_try_create (size_t size, htab_hash hash_f, htab_eq eq_f, htab_del del_f) |
045b3a49 GK |
388 | { |
389 | return htab_create_alloc (size, hash_f, eq_f, del_f, calloc, free); | |
390 | } | |
391 | ||
a2f945c6 VM |
392 | /* This function frees all memory allocated for given hash table. |
393 | Naturally the hash table must already exist. */ | |
394 | ||
395 | void | |
6da879de | 396 | htab_delete (htab_t htab) |
a2f945c6 | 397 | { |
d9175b87 RH |
398 | size_t size = htab_size (htab); |
399 | PTR *entries = htab->entries; | |
5dc9cffd | 400 | int i; |
e38992e8 | 401 | |
5dc9cffd | 402 | if (htab->del_f) |
d9175b87 RH |
403 | for (i = size - 1; i >= 0; i--) |
404 | if (entries[i] != EMPTY_ENTRY && entries[i] != DELETED_ENTRY) | |
405 | (*htab->del_f) (entries[i]); | |
5dc9cffd | 406 | |
e2500fed GK |
407 | if (htab->free_f != NULL) |
408 | { | |
d9175b87 | 409 | (*htab->free_f) (entries); |
e2500fed GK |
410 | (*htab->free_f) (htab); |
411 | } | |
74828682 DJ |
412 | else if (htab->free_with_arg_f != NULL) |
413 | { | |
d9175b87 | 414 | (*htab->free_with_arg_f) (htab->alloc_arg, entries); |
74828682 DJ |
415 | (*htab->free_with_arg_f) (htab->alloc_arg, htab); |
416 | } | |
a2f945c6 VM |
417 | } |
418 | ||
419 | /* This function clears all entries in the given hash table. */ | |
420 | ||
421 | void | |
6da879de | 422 | htab_empty (htab_t htab) |
a2f945c6 | 423 | { |
d9175b87 RH |
424 | size_t size = htab_size (htab); |
425 | PTR *entries = htab->entries; | |
5dc9cffd | 426 | int i; |
e38992e8 | 427 | |
5dc9cffd | 428 | if (htab->del_f) |
d9175b87 RH |
429 | for (i = size - 1; i >= 0; i--) |
430 | if (entries[i] != EMPTY_ENTRY && entries[i] != DELETED_ENTRY) | |
431 | (*htab->del_f) (entries[i]); | |
5dc9cffd | 432 | |
d9175b87 | 433 | memset (entries, 0, size * sizeof (PTR)); |
a2f945c6 VM |
434 | } |
435 | ||
8c5d513f BS |
436 | /* Similar to htab_find_slot, but without several unwanted side effects: |
437 | - Does not call htab->eq_f when it finds an existing entry. | |
438 | - Does not change the count of elements/searches/collisions in the | |
439 | hash table. | |
440 | This function also assumes there are no deleted entries in the table. | |
441 | HASH is the hash value for the element to be inserted. */ | |
e38992e8 | 442 | |
35e9340f | 443 | static PTR * |
6da879de | 444 | find_empty_slot_for_expand (htab_t htab, hashval_t hash) |
8c5d513f | 445 | { |
d9175b87 RH |
446 | hashval_t index = htab_mod (hash, htab); |
447 | size_t size = htab_size (htab); | |
4fc4e478 RH |
448 | PTR *slot = htab->entries + index; |
449 | hashval_t hash2; | |
450 | ||
451 | if (*slot == EMPTY_ENTRY) | |
452 | return slot; | |
453 | else if (*slot == DELETED_ENTRY) | |
454 | abort (); | |
8c5d513f | 455 | |
d9175b87 | 456 | hash2 = htab_mod_m2 (hash, htab); |
8c5d513f BS |
457 | for (;;) |
458 | { | |
4fc4e478 RH |
459 | index += hash2; |
460 | if (index >= size) | |
461 | index -= size; | |
e38992e8 | 462 | |
4fc4e478 | 463 | slot = htab->entries + index; |
8c5d513f BS |
464 | if (*slot == EMPTY_ENTRY) |
465 | return slot; | |
e38992e8 | 466 | else if (*slot == DELETED_ENTRY) |
8c5d513f | 467 | abort (); |
8c5d513f BS |
468 | } |
469 | } | |
470 | ||
a2f945c6 VM |
471 | /* The following function changes size of memory allocated for the |
472 | entries and repeatedly inserts the table elements. The occupancy | |
473 | of the table after the call will be about 50%. Naturally the hash | |
474 | table must already exist. Remember also that the place of the | |
d50d20ec HPN |
475 | table entries is changed. If memory allocation failures are allowed, |
476 | this function will return zero, indicating that the table could not be | |
477 | expanded. If all goes well, it will return a non-zero value. */ | |
a2f945c6 | 478 | |
d50d20ec | 479 | static int |
6da879de | 480 | htab_expand (htab_t htab) |
a2f945c6 | 481 | { |
35e9340f HPN |
482 | PTR *oentries; |
483 | PTR *olimit; | |
484 | PTR *p; | |
e2500fed | 485 | PTR *nentries; |
9bf3c9cc RH |
486 | size_t nsize, osize, elts; |
487 | unsigned int oindex, nindex; | |
5194cf08 ZW |
488 | |
489 | oentries = htab->entries; | |
9bf3c9cc RH |
490 | oindex = htab->size_prime_index; |
491 | osize = htab->size; | |
492 | olimit = oentries + osize; | |
493 | elts = htab_elements (htab); | |
5194cf08 | 494 | |
0a8e3de3 JH |
495 | /* Resize only when table after removal of unused elements is either |
496 | too full or too empty. */ | |
9bf3c9cc RH |
497 | if (elts * 2 > osize || (elts * 8 < osize && osize > 32)) |
498 | { | |
499 | nindex = higher_prime_index (elts * 2); | |
500 | nsize = prime_tab[nindex].prime; | |
501 | } | |
0a8e3de3 | 502 | else |
9bf3c9cc RH |
503 | { |
504 | nindex = oindex; | |
505 | nsize = osize; | |
506 | } | |
d50d20ec | 507 | |
74828682 DJ |
508 | if (htab->alloc_with_arg_f != NULL) |
509 | nentries = (PTR *) (*htab->alloc_with_arg_f) (htab->alloc_arg, nsize, | |
510 | sizeof (PTR *)); | |
511 | else | |
512 | nentries = (PTR *) (*htab->alloc_f) (nsize, sizeof (PTR *)); | |
e2500fed GK |
513 | if (nentries == NULL) |
514 | return 0; | |
515 | htab->entries = nentries; | |
120cdf68 | 516 | htab->size = nsize; |
9bf3c9cc | 517 | htab->size_prime_index = nindex; |
5194cf08 ZW |
518 | htab->n_elements -= htab->n_deleted; |
519 | htab->n_deleted = 0; | |
520 | ||
521 | p = oentries; | |
522 | do | |
523 | { | |
35e9340f | 524 | PTR x = *p; |
e38992e8 | 525 | |
5194cf08 ZW |
526 | if (x != EMPTY_ENTRY && x != DELETED_ENTRY) |
527 | { | |
35e9340f | 528 | PTR *q = find_empty_slot_for_expand (htab, (*htab->hash_f) (x)); |
e38992e8 | 529 | |
5194cf08 ZW |
530 | *q = x; |
531 | } | |
e38992e8 | 532 | |
5194cf08 ZW |
533 | p++; |
534 | } | |
535 | while (p < olimit); | |
e38992e8 | 536 | |
e2500fed GK |
537 | if (htab->free_f != NULL) |
538 | (*htab->free_f) (oentries); | |
74828682 DJ |
539 | else if (htab->free_with_arg_f != NULL) |
540 | (*htab->free_with_arg_f) (htab->alloc_arg, oentries); | |
d50d20ec | 541 | return 1; |
a2f945c6 VM |
542 | } |
543 | ||
5194cf08 ZW |
544 | /* This function searches for a hash table entry equal to the given |
545 | element. It cannot be used to insert or delete an element. */ | |
546 | ||
35e9340f | 547 | PTR |
6da879de | 548 | htab_find_with_hash (htab_t htab, const PTR element, hashval_t hash) |
a2f945c6 | 549 | { |
d9175b87 | 550 | hashval_t index, hash2; |
5194cf08 | 551 | size_t size; |
35e9340f | 552 | PTR entry; |
5194cf08 ZW |
553 | |
554 | htab->searches++; | |
d9175b87 RH |
555 | size = htab_size (htab); |
556 | index = htab_mod (hash, htab); | |
a2f945c6 | 557 | |
0194e877 ZW |
558 | entry = htab->entries[index]; |
559 | if (entry == EMPTY_ENTRY | |
560 | || (entry != DELETED_ENTRY && (*htab->eq_f) (entry, element))) | |
561 | return entry; | |
562 | ||
d9175b87 | 563 | hash2 = htab_mod_m2 (hash, htab); |
5194cf08 | 564 | for (;;) |
a2f945c6 | 565 | { |
5194cf08 ZW |
566 | htab->collisions++; |
567 | index += hash2; | |
568 | if (index >= size) | |
569 | index -= size; | |
0194e877 ZW |
570 | |
571 | entry = htab->entries[index]; | |
572 | if (entry == EMPTY_ENTRY | |
573 | || (entry != DELETED_ENTRY && (*htab->eq_f) (entry, element))) | |
574 | return entry; | |
a2f945c6 | 575 | } |
5194cf08 ZW |
576 | } |
577 | ||
8c5d513f BS |
578 | /* Like htab_find_slot_with_hash, but compute the hash value from the |
579 | element. */ | |
e38992e8 | 580 | |
35e9340f | 581 | PTR |
6da879de | 582 | htab_find (htab_t htab, const PTR element) |
8c5d513f BS |
583 | { |
584 | return htab_find_with_hash (htab, element, (*htab->hash_f) (element)); | |
585 | } | |
586 | ||
5194cf08 ZW |
587 | /* This function searches for a hash table slot containing an entry |
588 | equal to the given element. To delete an entry, call this with | |
6a88516c BE |
589 | insert=NO_INSERT, then call htab_clear_slot on the slot returned |
590 | (possibly after doing some checks). To insert an entry, call this | |
591 | with insert=INSERT, then write the value you want into the returned | |
592 | slot. When inserting an entry, NULL may be returned if memory | |
593 | allocation fails. */ | |
5194cf08 | 594 | |
35e9340f | 595 | PTR * |
6da879de GDR |
596 | htab_find_slot_with_hash (htab_t htab, const PTR element, |
597 | hashval_t hash, enum insert_option insert) | |
5194cf08 | 598 | { |
35e9340f | 599 | PTR *first_deleted_slot; |
d9175b87 | 600 | hashval_t index, hash2; |
5194cf08 | 601 | size_t size; |
4fc4e478 | 602 | PTR entry; |
5194cf08 | 603 | |
d9175b87 RH |
604 | size = htab_size (htab); |
605 | if (insert == INSERT && size * 3 <= htab->n_elements * 4) | |
606 | { | |
607 | if (htab_expand (htab) == 0) | |
608 | return NULL; | |
609 | size = htab_size (htab); | |
610 | } | |
5194cf08 | 611 | |
d9175b87 | 612 | index = htab_mod (hash, htab); |
5194cf08 | 613 | |
a2f945c6 | 614 | htab->searches++; |
5194cf08 ZW |
615 | first_deleted_slot = NULL; |
616 | ||
4fc4e478 RH |
617 | entry = htab->entries[index]; |
618 | if (entry == EMPTY_ENTRY) | |
619 | goto empty_entry; | |
620 | else if (entry == DELETED_ENTRY) | |
621 | first_deleted_slot = &htab->entries[index]; | |
622 | else if ((*htab->eq_f) (entry, element)) | |
623 | return &htab->entries[index]; | |
624 | ||
d9175b87 | 625 | hash2 = htab_mod_m2 (hash, htab); |
5194cf08 | 626 | for (;;) |
a2f945c6 | 627 | { |
4fc4e478 RH |
628 | htab->collisions++; |
629 | index += hash2; | |
630 | if (index >= size) | |
631 | index -= size; | |
632 | ||
633 | entry = htab->entries[index]; | |
5194cf08 | 634 | if (entry == EMPTY_ENTRY) |
4fc4e478 RH |
635 | goto empty_entry; |
636 | else if (entry == DELETED_ENTRY) | |
5194cf08 ZW |
637 | { |
638 | if (!first_deleted_slot) | |
639 | first_deleted_slot = &htab->entries[index]; | |
640 | } | |
4fc4e478 | 641 | else if ((*htab->eq_f) (entry, element)) |
e38992e8 | 642 | return &htab->entries[index]; |
a2f945c6 | 643 | } |
4fc4e478 RH |
644 | |
645 | empty_entry: | |
646 | if (insert == NO_INSERT) | |
647 | return NULL; | |
648 | ||
4fc4e478 RH |
649 | if (first_deleted_slot) |
650 | { | |
e0432c1c | 651 | htab->n_deleted--; |
4fc4e478 RH |
652 | *first_deleted_slot = EMPTY_ENTRY; |
653 | return first_deleted_slot; | |
654 | } | |
655 | ||
e0432c1c | 656 | htab->n_elements++; |
4fc4e478 | 657 | return &htab->entries[index]; |
a2f945c6 VM |
658 | } |
659 | ||
8c5d513f BS |
660 | /* Like htab_find_slot_with_hash, but compute the hash value from the |
661 | element. */ | |
e38992e8 | 662 | |
35e9340f | 663 | PTR * |
6da879de | 664 | htab_find_slot (htab_t htab, const PTR element, enum insert_option insert) |
8c5d513f BS |
665 | { |
666 | return htab_find_slot_with_hash (htab, element, (*htab->hash_f) (element), | |
667 | insert); | |
668 | } | |
669 | ||
7f96816a JL |
670 | /* This function deletes an element with the given value from hash |
671 | table (the hash is computed from the element). If there is no matching | |
672 | element in the hash table, this function does nothing. */ | |
673 | ||
674 | void | |
6da879de | 675 | htab_remove_elt (htab_t htab, PTR element) |
7f96816a JL |
676 | { |
677 | htab_remove_elt_with_hash (htab, element, (*htab->hash_f) (element)); | |
678 | } | |
679 | ||
680 | ||
5194cf08 ZW |
681 | /* This function deletes an element with the given value from hash |
682 | table. If there is no matching element in the hash table, this | |
683 | function does nothing. */ | |
a2f945c6 VM |
684 | |
685 | void | |
6da879de | 686 | htab_remove_elt_with_hash (htab_t htab, PTR element, hashval_t hash) |
a2f945c6 | 687 | { |
35e9340f | 688 | PTR *slot; |
a2f945c6 | 689 | |
7f96816a | 690 | slot = htab_find_slot_with_hash (htab, element, hash, NO_INSERT); |
5194cf08 ZW |
691 | if (*slot == EMPTY_ENTRY) |
692 | return; | |
693 | ||
5dc9cffd ZW |
694 | if (htab->del_f) |
695 | (*htab->del_f) (*slot); | |
696 | ||
5194cf08 ZW |
697 | *slot = DELETED_ENTRY; |
698 | htab->n_deleted++; | |
a2f945c6 VM |
699 | } |
700 | ||
5194cf08 ZW |
701 | /* This function clears a specified slot in a hash table. It is |
702 | useful when you've already done the lookup and don't want to do it | |
703 | again. */ | |
ed38f5d5 ZW |
704 | |
705 | void | |
6da879de | 706 | htab_clear_slot (htab_t htab, PTR *slot) |
ed38f5d5 | 707 | { |
d9175b87 | 708 | if (slot < htab->entries || slot >= htab->entries + htab_size (htab) |
ed38f5d5 ZW |
709 | || *slot == EMPTY_ENTRY || *slot == DELETED_ENTRY) |
710 | abort (); | |
e38992e8 | 711 | |
5dc9cffd ZW |
712 | if (htab->del_f) |
713 | (*htab->del_f) (*slot); | |
e38992e8 | 714 | |
ed38f5d5 | 715 | *slot = DELETED_ENTRY; |
5194cf08 | 716 | htab->n_deleted++; |
ed38f5d5 ZW |
717 | } |
718 | ||
719 | /* This function scans over the entire hash table calling | |
720 | CALLBACK for each live entry. If CALLBACK returns false, | |
721 | the iteration stops. INFO is passed as CALLBACK's second | |
722 | argument. */ | |
723 | ||
724 | void | |
6da879de | 725 | htab_traverse_noresize (htab_t htab, htab_trav callback, PTR info) |
ed38f5d5 | 726 | { |
0a8e3de3 JH |
727 | PTR *slot; |
728 | PTR *limit; | |
729 | ||
0a8e3de3 | 730 | slot = htab->entries; |
d9175b87 | 731 | limit = slot + htab_size (htab); |
e38992e8 | 732 | |
5194cf08 ZW |
733 | do |
734 | { | |
35e9340f | 735 | PTR x = *slot; |
e38992e8 | 736 | |
5194cf08 | 737 | if (x != EMPTY_ENTRY && x != DELETED_ENTRY) |
8c5d513f | 738 | if (!(*callback) (slot, info)) |
5194cf08 ZW |
739 | break; |
740 | } | |
741 | while (++slot < limit); | |
ed38f5d5 ZW |
742 | } |
743 | ||
dbccdc42 JH |
744 | /* Like htab_traverse_noresize, but does resize the table when it is |
745 | too empty to improve effectivity of subsequent calls. */ | |
746 | ||
747 | void | |
6da879de | 748 | htab_traverse (htab_t htab, htab_trav callback, PTR info) |
dbccdc42 | 749 | { |
d9175b87 | 750 | if (htab_elements (htab) * 8 < htab_size (htab)) |
dbccdc42 JH |
751 | htab_expand (htab); |
752 | ||
753 | htab_traverse_noresize (htab, callback, info); | |
754 | } | |
755 | ||
e38992e8 RK |
756 | /* Return the fraction of fixed collisions during all work with given |
757 | hash table. */ | |
a2f945c6 | 758 | |
5194cf08 | 759 | double |
6da879de | 760 | htab_collisions (htab_t htab) |
a2f945c6 | 761 | { |
e38992e8 | 762 | if (htab->searches == 0) |
5194cf08 | 763 | return 0.0; |
e38992e8 RK |
764 | |
765 | return (double) htab->collisions / (double) htab->searches; | |
a2f945c6 | 766 | } |
9e0ba685 | 767 | |
0ed5305d RH |
768 | /* Hash P as a null-terminated string. |
769 | ||
770 | Copied from gcc/hashtable.c. Zack had the following to say with respect | |
771 | to applicability, though note that unlike hashtable.c, this hash table | |
772 | implementation re-hashes rather than chain buckets. | |
773 | ||
774 | http://gcc.gnu.org/ml/gcc-patches/2001-08/msg01021.html | |
775 | From: Zack Weinberg <zackw@panix.com> | |
776 | Date: Fri, 17 Aug 2001 02:15:56 -0400 | |
777 | ||
778 | I got it by extracting all the identifiers from all the source code | |
779 | I had lying around in mid-1999, and testing many recurrences of | |
780 | the form "H_n = H_{n-1} * K + c_n * L + M" where K, L, M were either | |
781 | prime numbers or the appropriate identity. This was the best one. | |
782 | I don't remember exactly what constituted "best", except I was | |
783 | looking at bucket-length distributions mostly. | |
784 | ||
785 | So it should be very good at hashing identifiers, but might not be | |
786 | as good at arbitrary strings. | |
787 | ||
788 | I'll add that it thoroughly trounces the hash functions recommended | |
789 | for this use at http://burtleburtle.net/bob/hash/index.html, both | |
790 | on speed and bucket distribution. I haven't tried it against the | |
791 | function they just started using for Perl's hashes. */ | |
9e0ba685 RH |
792 | |
793 | hashval_t | |
6da879de | 794 | htab_hash_string (const PTR p) |
9e0ba685 RH |
795 | { |
796 | const unsigned char *str = (const unsigned char *) p; | |
797 | hashval_t r = 0; | |
798 | unsigned char c; | |
799 | ||
800 | while ((c = *str++) != 0) | |
801 | r = r * 67 + c - 113; | |
802 | ||
803 | return r; | |
804 | } | |
5cc5a0d0 JM |
805 | |
806 | /* DERIVED FROM: | |
807 | -------------------------------------------------------------------- | |
808 | lookup2.c, by Bob Jenkins, December 1996, Public Domain. | |
809 | hash(), hash2(), hash3, and mix() are externally useful functions. | |
810 | Routines to test the hash are included if SELF_TEST is defined. | |
811 | You can use this free for any purpose. It has no warranty. | |
812 | -------------------------------------------------------------------- | |
813 | */ | |
814 | ||
815 | /* | |
816 | -------------------------------------------------------------------- | |
817 | mix -- mix 3 32-bit values reversibly. | |
818 | For every delta with one or two bit set, and the deltas of all three | |
819 | high bits or all three low bits, whether the original value of a,b,c | |
820 | is almost all zero or is uniformly distributed, | |
821 | * If mix() is run forward or backward, at least 32 bits in a,b,c | |
822 | have at least 1/4 probability of changing. | |
823 | * If mix() is run forward, every bit of c will change between 1/3 and | |
824 | 2/3 of the time. (Well, 22/100 and 78/100 for some 2-bit deltas.) | |
825 | mix() was built out of 36 single-cycle latency instructions in a | |
826 | structure that could supported 2x parallelism, like so: | |
827 | a -= b; | |
828 | a -= c; x = (c>>13); | |
829 | b -= c; a ^= x; | |
830 | b -= a; x = (a<<8); | |
831 | c -= a; b ^= x; | |
832 | c -= b; x = (b>>13); | |
833 | ... | |
834 | Unfortunately, superscalar Pentiums and Sparcs can't take advantage | |
835 | of that parallelism. They've also turned some of those single-cycle | |
836 | latency instructions into multi-cycle latency instructions. Still, | |
837 | this is the fastest good hash I could find. There were about 2^^68 | |
838 | to choose from. I only looked at a billion or so. | |
839 | -------------------------------------------------------------------- | |
840 | */ | |
841 | /* same, but slower, works on systems that might have 8 byte hashval_t's */ | |
842 | #define mix(a,b,c) \ | |
843 | { \ | |
844 | a -= b; a -= c; a ^= (c>>13); \ | |
845 | b -= c; b -= a; b ^= (a<< 8); \ | |
846 | c -= a; c -= b; c ^= ((b&0xffffffff)>>13); \ | |
847 | a -= b; a -= c; a ^= ((c&0xffffffff)>>12); \ | |
848 | b -= c; b -= a; b = (b ^ (a<<16)) & 0xffffffff; \ | |
849 | c -= a; c -= b; c = (c ^ (b>> 5)) & 0xffffffff; \ | |
850 | a -= b; a -= c; a = (a ^ (c>> 3)) & 0xffffffff; \ | |
851 | b -= c; b -= a; b = (b ^ (a<<10)) & 0xffffffff; \ | |
852 | c -= a; c -= b; c = (c ^ (b>>15)) & 0xffffffff; \ | |
853 | } | |
854 | ||
855 | /* | |
856 | -------------------------------------------------------------------- | |
857 | hash() -- hash a variable-length key into a 32-bit value | |
858 | k : the key (the unaligned variable-length array of bytes) | |
859 | len : the length of the key, counting by bytes | |
860 | level : can be any 4-byte value | |
861 | Returns a 32-bit value. Every bit of the key affects every bit of | |
862 | the return value. Every 1-bit and 2-bit delta achieves avalanche. | |
863 | About 36+6len instructions. | |
864 | ||
865 | The best hash table sizes are powers of 2. There is no need to do | |
866 | mod a prime (mod is sooo slow!). If you need less than 32 bits, | |
867 | use a bitmask. For example, if you need only 10 bits, do | |
868 | h = (h & hashmask(10)); | |
869 | In which case, the hash table should have hashsize(10) elements. | |
870 | ||
871 | If you are hashing n strings (ub1 **)k, do it like this: | |
872 | for (i=0, h=0; i<n; ++i) h = hash( k[i], len[i], h); | |
873 | ||
874 | By Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may use this | |
875 | code any way you wish, private, educational, or commercial. It's free. | |
876 | ||
877 | See http://burtleburtle.net/bob/hash/evahash.html | |
878 | Use for hash table lookup, or anything where one collision in 2^32 is | |
879 | acceptable. Do NOT use for cryptographic purposes. | |
880 | -------------------------------------------------------------------- | |
881 | */ | |
882 | ||
6da879de GDR |
883 | hashval_t |
884 | iterative_hash (const PTR k_in /* the key */, | |
885 | register size_t length /* the length of the key */, | |
886 | register hashval_t initval /* the previous hash, or | |
887 | an arbitrary value */) | |
5cc5a0d0 JM |
888 | { |
889 | register const unsigned char *k = (const unsigned char *)k_in; | |
890 | register hashval_t a,b,c,len; | |
891 | ||
892 | /* Set up the internal state */ | |
893 | len = length; | |
894 | a = b = 0x9e3779b9; /* the golden ratio; an arbitrary value */ | |
895 | c = initval; /* the previous hash value */ | |
896 | ||
897 | /*---------------------------------------- handle most of the key */ | |
898 | #ifndef WORDS_BIGENDIAN | |
899 | /* On a little-endian machine, if the data is 4-byte aligned we can hash | |
900 | by word for better speed. This gives nondeterministic results on | |
901 | big-endian machines. */ | |
902 | if (sizeof (hashval_t) == 4 && (((size_t)k)&3) == 0) | |
903 | while (len >= 12) /* aligned */ | |
904 | { | |
905 | a += *(hashval_t *)(k+0); | |
906 | b += *(hashval_t *)(k+4); | |
907 | c += *(hashval_t *)(k+8); | |
908 | mix(a,b,c); | |
909 | k += 12; len -= 12; | |
910 | } | |
911 | else /* unaligned */ | |
912 | #endif | |
913 | while (len >= 12) | |
914 | { | |
915 | a += (k[0] +((hashval_t)k[1]<<8) +((hashval_t)k[2]<<16) +((hashval_t)k[3]<<24)); | |
916 | b += (k[4] +((hashval_t)k[5]<<8) +((hashval_t)k[6]<<16) +((hashval_t)k[7]<<24)); | |
917 | c += (k[8] +((hashval_t)k[9]<<8) +((hashval_t)k[10]<<16)+((hashval_t)k[11]<<24)); | |
918 | mix(a,b,c); | |
919 | k += 12; len -= 12; | |
920 | } | |
921 | ||
922 | /*------------------------------------- handle the last 11 bytes */ | |
923 | c += length; | |
924 | switch(len) /* all the case statements fall through */ | |
925 | { | |
926 | case 11: c+=((hashval_t)k[10]<<24); | |
927 | case 10: c+=((hashval_t)k[9]<<16); | |
928 | case 9 : c+=((hashval_t)k[8]<<8); | |
929 | /* the first byte of c is reserved for the length */ | |
930 | case 8 : b+=((hashval_t)k[7]<<24); | |
931 | case 7 : b+=((hashval_t)k[6]<<16); | |
932 | case 6 : b+=((hashval_t)k[5]<<8); | |
933 | case 5 : b+=k[4]; | |
934 | case 4 : a+=((hashval_t)k[3]<<24); | |
935 | case 3 : a+=((hashval_t)k[2]<<16); | |
936 | case 2 : a+=((hashval_t)k[1]<<8); | |
937 | case 1 : a+=k[0]; | |
938 | /* case 0: nothing left to add */ | |
939 | } | |
940 | mix(a,b,c); | |
941 | /*-------------------------------------------- report the result */ | |
942 | return c; | |
943 | } |