Commit | Line | Data |
---|---|---|
c906108c | 1 | /* Implement a cached obstack. |
c2d11a7d JM |
2 | Written by Fred Fish <fnf@cygnus.com> |
3 | Rewritten by Jim Blandy <jimb@cygnus.com> | |
2c7ef074 AC |
4 | |
5 | Copyright 1999, 2000, 2002 Free Software Foundation, Inc. | |
c906108c | 6 | |
c5aa993b | 7 | This file is part of GDB. |
c906108c | 8 | |
c5aa993b JM |
9 | This program is free software; you can redistribute it and/or modify |
10 | it under the terms of the GNU General Public License as published by | |
11 | the Free Software Foundation; either version 2 of the License, or | |
12 | (at your option) any later version. | |
c906108c | 13 | |
c5aa993b JM |
14 | This program is distributed in the hope that it will be useful, |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
17 | GNU General Public License for more details. | |
c906108c | 18 | |
c5aa993b JM |
19 | You should have received a copy of the GNU General Public License |
20 | along with this program; if not, write to the Free Software | |
21 | Foundation, Inc., 59 Temple Place - Suite 330, | |
22 | Boston, MA 02111-1307, USA. */ | |
c906108c SS |
23 | |
24 | #include "defs.h" | |
25 | #include "obstack.h" | |
26 | #include "bcache.h" | |
27 | #include "gdb_string.h" /* For memcpy declaration */ | |
28 | ||
2c7ef074 AC |
29 | #include <stddef.h> |
30 | #include <stdlib.h> | |
31 | ||
af5f3db6 AC |
32 | /* The type used to hold a single bcache string. The user data is |
33 | stored in d.data. Since it can be any type, it needs to have the | |
34 | same alignment as the most strict alignment of any type on the host | |
35 | machine. I don't know of any really correct way to do this in | |
36 | stock ANSI C, so just do it the same way obstack.h does. */ | |
37 | ||
38 | struct bstring | |
39 | { | |
40 | struct bstring *next; | |
41 | size_t length; | |
42 | ||
43 | union | |
44 | { | |
45 | char data[1]; | |
46 | double dummy; | |
47 | } | |
48 | d; | |
49 | }; | |
50 | ||
51 | ||
52 | /* The structure for a bcache itself. The bcache is initialized, in | |
53 | bcache_xmalloc(), by filling it with zeros and then setting the | |
54 | corresponding obstack's malloc() and free() methods. */ | |
55 | ||
56 | struct bcache | |
57 | { | |
58 | /* All the bstrings are allocated here. */ | |
59 | struct obstack cache; | |
60 | ||
61 | /* How many hash buckets we're using. */ | |
62 | unsigned int num_buckets; | |
63 | ||
64 | /* Hash buckets. This table is allocated using malloc, so when we | |
65 | grow the table we can return the old table to the system. */ | |
66 | struct bstring **bucket; | |
67 | ||
68 | /* Statistics. */ | |
69 | unsigned long unique_count; /* number of unique strings */ | |
70 | long total_count; /* total number of strings cached, including dups */ | |
71 | long unique_size; /* size of unique strings, in bytes */ | |
72 | long total_size; /* total number of bytes cached, including dups */ | |
73 | long structure_size; /* total size of bcache, including infrastructure */ | |
74 | }; | |
75 | ||
357e46e7 DB |
76 | /* The old hash function was stolen from SDBM. This is what DB 3.0 uses now, |
77 | * and is better than the old one. | |
78 | */ | |
c2d11a7d | 79 | \f |
c2d11a7d | 80 | unsigned long |
d85a5daf | 81 | hash(const void *addr, int length) |
c2d11a7d | 82 | { |
357e46e7 DB |
83 | const unsigned char *k, *e; |
84 | unsigned long h; | |
85 | ||
86 | k = (const unsigned char *)addr; | |
87 | e = k+length; | |
88 | for (h=0; k< e;++k) | |
89 | { | |
90 | h *=16777619; | |
91 | h ^= *k; | |
92 | } | |
93 | return (h); | |
c906108c | 94 | } |
c2d11a7d JM |
95 | \f |
96 | /* Growing the bcache's hash table. */ | |
97 | ||
98 | /* If the average chain length grows beyond this, then we want to | |
99 | resize our hash table. */ | |
100 | #define CHAIN_LENGTH_THRESHOLD (5) | |
101 | ||
102 | static void | |
103 | expand_hash_table (struct bcache *bcache) | |
c906108c | 104 | { |
c2d11a7d JM |
105 | /* A table of good hash table sizes. Whenever we grow, we pick the |
106 | next larger size from this table. sizes[i] is close to 1 << (i+10), | |
107 | so we roughly double the table size each time. After we fall off | |
108 | the end of this table, we just double. Don't laugh --- there have | |
109 | been executables sighted with a gigabyte of debug info. */ | |
110 | static unsigned long sizes[] = { | |
111 | 1021, 2053, 4099, 8191, 16381, 32771, | |
112 | 65537, 131071, 262144, 524287, 1048573, 2097143, | |
113 | 4194301, 8388617, 16777213, 33554467, 67108859, 134217757, | |
114 | 268435459, 536870923, 1073741827, 2147483659UL | |
115 | }; | |
745b8ca0 | 116 | unsigned int new_num_buckets; |
c2d11a7d | 117 | struct bstring **new_buckets; |
745b8ca0 | 118 | unsigned int i; |
c2d11a7d JM |
119 | |
120 | /* Find the next size. */ | |
745b8ca0 | 121 | new_num_buckets = bcache->num_buckets * 2; |
c2d11a7d JM |
122 | for (i = 0; i < (sizeof (sizes) / sizeof (sizes[0])); i++) |
123 | if (sizes[i] > bcache->num_buckets) | |
124 | { | |
125 | new_num_buckets = sizes[i]; | |
126 | break; | |
127 | } | |
c2d11a7d JM |
128 | |
129 | /* Allocate the new table. */ | |
130 | { | |
131 | size_t new_size = new_num_buckets * sizeof (new_buckets[0]); | |
132 | new_buckets = (struct bstring **) xmalloc (new_size); | |
133 | memset (new_buckets, 0, new_size); | |
134 | ||
135 | bcache->structure_size -= (bcache->num_buckets | |
136 | * sizeof (bcache->bucket[0])); | |
137 | bcache->structure_size += new_size; | |
138 | } | |
c906108c | 139 | |
c2d11a7d JM |
140 | /* Rehash all existing strings. */ |
141 | for (i = 0; i < bcache->num_buckets; i++) | |
c906108c | 142 | { |
c2d11a7d JM |
143 | struct bstring *s, *next; |
144 | ||
145 | for (s = bcache->bucket[i]; s; s = next) | |
c906108c | 146 | { |
c2d11a7d JM |
147 | struct bstring **new_bucket; |
148 | next = s->next; | |
149 | ||
150 | new_bucket = &new_buckets[(hash (&s->d.data, s->length) | |
151 | % new_num_buckets)]; | |
152 | s->next = *new_bucket; | |
153 | *new_bucket = s; | |
c906108c SS |
154 | } |
155 | } | |
c2d11a7d JM |
156 | |
157 | /* Plug in the new table. */ | |
158 | if (bcache->bucket) | |
b8c9b27d | 159 | xfree (bcache->bucket); |
c2d11a7d JM |
160 | bcache->bucket = new_buckets; |
161 | bcache->num_buckets = new_num_buckets; | |
c906108c SS |
162 | } |
163 | ||
c2d11a7d JM |
164 | \f |
165 | /* Looking up things in the bcache. */ | |
166 | ||
167 | /* The number of bytes needed to allocate a struct bstring whose data | |
168 | is N bytes long. */ | |
169 | #define BSTRING_SIZE(n) (offsetof (struct bstring, d.data) + (n)) | |
170 | ||
171 | /* Find a copy of the LENGTH bytes at ADDR in BCACHE. If BCACHE has | |
172 | never seen those bytes before, add a copy of them to BCACHE. In | |
173 | either case, return a pointer to BCACHE's copy of that string. */ | |
c906108c | 174 | void * |
d85a5daf | 175 | bcache (const void *addr, int length, struct bcache *bcache) |
c906108c | 176 | { |
c2d11a7d JM |
177 | int hash_index; |
178 | struct bstring *s; | |
c906108c | 179 | |
c2d11a7d JM |
180 | /* If our average chain length is too high, expand the hash table. */ |
181 | if (bcache->unique_count >= bcache->num_buckets * CHAIN_LENGTH_THRESHOLD) | |
182 | expand_hash_table (bcache); | |
183 | ||
184 | bcache->total_count++; | |
185 | bcache->total_size += length; | |
186 | ||
187 | hash_index = hash (addr, length) % bcache->num_buckets; | |
188 | ||
189 | /* Search the hash bucket for a string identical to the caller's. */ | |
190 | for (s = bcache->bucket[hash_index]; s; s = s->next) | |
191 | if (s->length == length | |
192 | && ! memcmp (&s->d.data, addr, length)) | |
193 | return &s->d.data; | |
194 | ||
195 | /* The user's string isn't in the list. Insert it after *ps. */ | |
196 | { | |
197 | struct bstring *new | |
198 | = obstack_alloc (&bcache->cache, BSTRING_SIZE (length)); | |
199 | memcpy (&new->d.data, addr, length); | |
200 | new->length = length; | |
201 | new->next = bcache->bucket[hash_index]; | |
202 | bcache->bucket[hash_index] = new; | |
203 | ||
204 | bcache->unique_count++; | |
205 | bcache->unique_size += length; | |
206 | bcache->structure_size += BSTRING_SIZE (length); | |
207 | ||
208 | return &new->d.data; | |
209 | } | |
c906108c SS |
210 | } |
211 | ||
c2d11a7d | 212 | \f |
af5f3db6 AC |
213 | /* Allocating and freeing bcaches. */ |
214 | ||
215 | struct bcache * | |
216 | bcache_xmalloc (void) | |
217 | { | |
218 | /* Allocate the bcache pre-zeroed. */ | |
219 | struct bcache *b = XCALLOC (1, struct bcache); | |
220 | obstack_specify_allocation (&b->cache, 0, 0, xmalloc, xfree); | |
221 | return b; | |
222 | } | |
c2d11a7d JM |
223 | |
224 | /* Free all the storage associated with BCACHE. */ | |
c906108c | 225 | void |
af5f3db6 | 226 | bcache_xfree (struct bcache *bcache) |
c906108c | 227 | { |
af5f3db6 AC |
228 | if (bcache == NULL) |
229 | return; | |
c2d11a7d | 230 | obstack_free (&bcache->cache, 0); |
af5f3db6 AC |
231 | xfree (bcache->bucket); |
232 | xfree (bcache); | |
c2d11a7d JM |
233 | } |
234 | ||
235 | ||
236 | \f | |
237 | /* Printing statistics. */ | |
238 | ||
239 | static int | |
240 | compare_ints (const void *ap, const void *bp) | |
241 | { | |
242 | /* Because we know we're comparing two ints which are positive, | |
243 | there's no danger of overflow here. */ | |
244 | return * (int *) ap - * (int *) bp; | |
245 | } | |
246 | ||
247 | ||
248 | static void | |
249 | print_percentage (int portion, int total) | |
250 | { | |
251 | if (total == 0) | |
252 | printf_filtered ("(not applicable)\n"); | |
c906108c | 253 | else |
c2d11a7d JM |
254 | printf_filtered ("%3d%%\n", portion * 100 / total); |
255 | } | |
256 | ||
257 | ||
258 | /* Print statistics on BCACHE's memory usage and efficacity at | |
259 | eliminating duplication. NAME should describe the kind of data | |
260 | BCACHE holds. Statistics are printed using `printf_filtered' and | |
261 | its ilk. */ | |
262 | void | |
263 | print_bcache_statistics (struct bcache *c, char *type) | |
264 | { | |
265 | int occupied_buckets; | |
266 | int max_chain_length; | |
267 | int median_chain_length; | |
268 | ||
269 | /* Count the number of occupied buckets, and measure chain lengths. */ | |
270 | { | |
745b8ca0 | 271 | unsigned int b; |
c2d11a7d JM |
272 | int *chain_length |
273 | = (int *) alloca (c->num_buckets * sizeof (*chain_length)); | |
274 | ||
275 | occupied_buckets = 0; | |
276 | ||
277 | for (b = 0; b < c->num_buckets; b++) | |
278 | { | |
279 | struct bstring *s = c->bucket[b]; | |
280 | ||
281 | chain_length[b] = 0; | |
282 | ||
283 | if (s) | |
284 | { | |
285 | occupied_buckets++; | |
286 | ||
287 | while (s) | |
288 | { | |
289 | chain_length[b]++; | |
290 | s = s->next; | |
291 | } | |
292 | } | |
293 | } | |
294 | ||
295 | /* To compute the median, we need the set of chain lengths sorted. */ | |
296 | qsort (chain_length, c->num_buckets, sizeof (chain_length[0]), | |
297 | compare_ints); | |
298 | ||
299 | if (c->num_buckets > 0) | |
300 | { | |
301 | max_chain_length = chain_length[c->num_buckets - 1]; | |
302 | median_chain_length = chain_length[c->num_buckets / 2]; | |
303 | } | |
304 | else | |
305 | { | |
306 | max_chain_length = 0; | |
307 | median_chain_length = 0; | |
308 | } | |
309 | } | |
310 | ||
311 | printf_filtered (" Cached '%s' statistics:\n", type); | |
312 | printf_filtered (" Total object count: %ld\n", c->total_count); | |
745b8ca0 | 313 | printf_filtered (" Unique object count: %lu\n", c->unique_count); |
c2d11a7d JM |
314 | printf_filtered (" Percentage of duplicates, by count: "); |
315 | print_percentage (c->total_count - c->unique_count, c->total_count); | |
316 | printf_filtered ("\n"); | |
317 | ||
318 | printf_filtered (" Total object size: %ld\n", c->total_size); | |
319 | printf_filtered (" Unique object size: %ld\n", c->unique_size); | |
320 | printf_filtered (" Percentage of duplicates, by size: "); | |
321 | print_percentage (c->total_size - c->unique_size, c->total_size); | |
322 | printf_filtered ("\n"); | |
323 | ||
324 | printf_filtered (" Total memory used by bcache, including overhead: %ld\n", | |
325 | c->structure_size); | |
326 | printf_filtered (" Percentage memory overhead: "); | |
327 | print_percentage (c->structure_size - c->unique_size, c->unique_size); | |
328 | printf_filtered (" Net memory savings: "); | |
329 | print_percentage (c->total_size - c->structure_size, c->total_size); | |
330 | printf_filtered ("\n"); | |
331 | ||
332 | printf_filtered (" Hash table size: %3d\n", c->num_buckets); | |
333 | printf_filtered (" Hash table population: "); | |
334 | print_percentage (occupied_buckets, c->num_buckets); | |
335 | printf_filtered (" Median hash chain length: %3d\n", | |
336 | median_chain_length); | |
337 | printf_filtered (" Average hash chain length: "); | |
338 | if (c->num_buckets > 0) | |
745b8ca0 | 339 | printf_filtered ("%3lu\n", c->unique_count / c->num_buckets); |
c906108c | 340 | else |
c2d11a7d JM |
341 | printf_filtered ("(not applicable)\n"); |
342 | printf_filtered (" Maximum hash chain length: %3d\n", max_chain_length); | |
343 | printf_filtered ("\n"); | |
c906108c | 344 | } |
af5f3db6 AC |
345 | |
346 | int | |
347 | bcache_memory_used (struct bcache *bcache) | |
348 | { | |
349 | return obstack_memory_used (&bcache->cache); | |
350 | } |