Mercurial > hg > CbC > CbC_gcc
annotate gcc/ggc-page.c @ 144:8f4e72ab4e11
fix segmentation fault caused by nothing next cur_op to end
author | Takahiro SHIMIZU <anatofuz@cr.ie.u-ryukyu.ac.jp> |
---|---|
date | Sun, 23 Dec 2018 21:23:56 +0900 |
parents | 84e7813d76e9 |
children | 1830386684a0 |
rev | line source |
---|---|
0 | 1 /* "Bag-of-pages" garbage collector for the GNU compiler. |
131 | 2 Copyright (C) 1999-2018 Free Software Foundation, Inc. |
0 | 3 |
4 This file is part of GCC. | |
5 | |
6 GCC is free software; you can redistribute it and/or modify it under | |
7 the terms of the GNU General Public License as published by the Free | |
8 Software Foundation; either version 3, or (at your option) any later | |
9 version. | |
10 | |
11 GCC is distributed in the hope that it will be useful, but WITHOUT ANY | |
12 WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
14 for more details. | |
15 | |
16 You should have received a copy of the GNU General Public License | |
17 along with GCC; see the file COPYING3. If not see | |
18 <http://www.gnu.org/licenses/>. */ | |
19 | |
20 #include "config.h" | |
21 #include "system.h" | |
22 #include "coretypes.h" | |
111 | 23 #include "backend.h" |
24 #include "alias.h" | |
0 | 25 #include "tree.h" |
26 #include "rtl.h" | |
111 | 27 #include "memmodel.h" |
0 | 28 #include "tm_p.h" |
67
f6334be47118
update gcc from gcc-4.6-20100522 to gcc-4.6-20110318
nobuyasu <dimolto@cr.ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
29 #include "diagnostic-core.h" |
0 | 30 #include "flags.h" |
67
f6334be47118
update gcc from gcc-4.6-20100522 to gcc-4.6-20110318
nobuyasu <dimolto@cr.ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
31 #include "ggc-internal.h" |
0 | 32 #include "timevar.h" |
33 #include "params.h" | |
111 | 34 #include "cgraph.h" |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
35 #include "cfgloop.h" |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
36 #include "plugin.h" |
0 | 37 |
38 /* Prefer MAP_ANON(YMOUS) to /dev/zero, since we don't need to keep a | |
39 file open. Prefer either to valloc. */ | |
40 #ifdef HAVE_MMAP_ANON | |
41 # undef HAVE_MMAP_DEV_ZERO | |
42 # define USING_MMAP | |
43 #endif | |
44 | |
45 #ifdef HAVE_MMAP_DEV_ZERO | |
46 # define USING_MMAP | |
47 #endif | |
48 | |
49 #ifndef USING_MMAP | |
50 #define USING_MALLOC_PAGE_GROUPS | |
51 #endif | |
52 | |
111 | 53 #if defined(HAVE_MADVISE) && HAVE_DECL_MADVISE && defined(MADV_DONTNEED) \ |
54 && defined(USING_MMAP) | |
55 # define USING_MADVISE | |
56 #endif | |
57 | |
0 | 58 /* Strategy: |
59 | |
60 This garbage-collecting allocator allocates objects on one of a set | |
61 of pages. Each page can allocate objects of a single size only; | |
62 available sizes are powers of two starting at four bytes. The size | |
63 of an allocation request is rounded up to the next power of two | |
64 (`order'), and satisfied from the appropriate page. | |
65 | |
66 Each page is recorded in a page-entry, which also maintains an | |
67 in-use bitmap of object positions on the page. This allows the | |
68 allocation state of a particular object to be flipped without | |
69 touching the page itself. | |
70 | |
71 Each page-entry also has a context depth, which is used to track | |
72 pushing and popping of allocation contexts. Only objects allocated | |
73 in the current (highest-numbered) context may be collected. | |
74 | |
75 Page entries are arranged in an array of singly-linked lists. The | |
76 array is indexed by the allocation size, in bits, of the pages on | |
77 it; i.e. all pages on a list allocate objects of the same size. | |
78 Pages are ordered on the list such that all non-full pages precede | |
79 all full pages, with non-full pages arranged in order of decreasing | |
80 context depth. | |
81 | |
82 Empty pages (of all orders) are kept on a single page cache list, | |
83 and are considered first when new pages are required; they are | |
84 deallocated at the start of the next collection if they haven't | |
85 been recycled by then. */ | |
86 | |
87 /* Define GGC_DEBUG_LEVEL to print debugging information. | |
88 0: No debugging output. | |
89 1: GC statistics only. | |
90 2: Page-entry allocations/deallocations as well. | |
91 3: Object allocations as well. | |
92 4: Object marks as well. */ | |
93 #define GGC_DEBUG_LEVEL (0) | |
94 | |
95 /* A two-level tree is used to look up the page-entry for a given | |
96 pointer. Two chunks of the pointer's bits are extracted to index | |
97 the first and second levels of the tree, as follows: | |
98 | |
99 HOST_PAGE_SIZE_BITS | |
100 32 | | | |
101 msb +----------------+----+------+------+ lsb | |
102 | | | | |
103 PAGE_L1_BITS | | |
104 | | | |
105 PAGE_L2_BITS | |
106 | |
107 The bottommost HOST_PAGE_SIZE_BITS are ignored, since page-entry | |
108 pages are aligned on system page boundaries. The next most | |
109 significant PAGE_L2_BITS and PAGE_L1_BITS are the second and first | |
110 index values in the lookup table, respectively. | |
111 | |
112 For 32-bit architectures and the settings below, there are no | |
113 leftover bits. For architectures with wider pointers, the lookup | |
114 tree points to a list of pages, which must be scanned to find the | |
115 correct one. */ | |
116 | |
117 #define PAGE_L1_BITS (8) | |
118 #define PAGE_L2_BITS (32 - PAGE_L1_BITS - G.lg_pagesize) | |
111 | 119 #define PAGE_L1_SIZE ((uintptr_t) 1 << PAGE_L1_BITS) |
120 #define PAGE_L2_SIZE ((uintptr_t) 1 << PAGE_L2_BITS) | |
0 | 121 |
122 #define LOOKUP_L1(p) \ | |
111 | 123 (((uintptr_t) (p) >> (32 - PAGE_L1_BITS)) & ((1 << PAGE_L1_BITS) - 1)) |
0 | 124 |
125 #define LOOKUP_L2(p) \ | |
111 | 126 (((uintptr_t) (p) >> G.lg_pagesize) & ((1 << PAGE_L2_BITS) - 1)) |
0 | 127 |
128 /* The number of objects per allocation page, for objects on a page of | |
129 the indicated ORDER. */ | |
130 #define OBJECTS_PER_PAGE(ORDER) objects_per_page_table[ORDER] | |
131 | |
132 /* The number of objects in P. */ | |
133 #define OBJECTS_IN_PAGE(P) ((P)->bytes / OBJECT_SIZE ((P)->order)) | |
134 | |
135 /* The size of an object on a page of the indicated ORDER. */ | |
136 #define OBJECT_SIZE(ORDER) object_size_table[ORDER] | |
137 | |
138 /* For speed, we avoid doing a general integer divide to locate the | |
139 offset in the allocation bitmap, by precalculating numbers M, S | |
140 such that (O * M) >> S == O / Z (modulo 2^32), for any offset O | |
141 within the page which is evenly divisible by the object size Z. */ | |
142 #define DIV_MULT(ORDER) inverse_table[ORDER].mult | |
143 #define DIV_SHIFT(ORDER) inverse_table[ORDER].shift | |
144 #define OFFSET_TO_BIT(OFFSET, ORDER) \ | |
145 (((OFFSET) * DIV_MULT (ORDER)) >> DIV_SHIFT (ORDER)) | |
146 | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
147 /* We use this structure to determine the alignment required for |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
148 allocations. For power-of-two sized allocations, that's not a |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
149 problem, but it does matter for odd-sized allocations. |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
150 We do not care about alignment for floating-point types. */ |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
151 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
152 struct max_alignment { |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
153 char c; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
154 union { |
111 | 155 int64_t i; |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
156 void *p; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
157 } u; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
158 }; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
159 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
160 /* The biggest alignment required. */ |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
161 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
162 #define MAX_ALIGNMENT (offsetof (struct max_alignment, u)) |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
163 |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
164 |
0 | 165 /* The number of extra orders, not corresponding to power-of-two sized |
166 objects. */ | |
167 | |
168 #define NUM_EXTRA_ORDERS ARRAY_SIZE (extra_order_size_table) | |
169 | |
170 #define RTL_SIZE(NSLOTS) \ | |
171 (RTX_HDR_SIZE + (NSLOTS) * sizeof (rtunion)) | |
172 | |
173 #define TREE_EXP_SIZE(OPS) \ | |
174 (sizeof (struct tree_exp) + ((OPS) - 1) * sizeof (tree)) | |
175 | |
176 /* The Ith entry is the maximum size of an object to be stored in the | |
177 Ith extra order. Adding a new entry to this array is the *only* | |
178 thing you need to do to add a new special allocation size. */ | |
179 | |
180 static const size_t extra_order_size_table[] = { | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
181 /* Extra orders for small non-power-of-two multiples of MAX_ALIGNMENT. |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
182 There are a lot of structures with these sizes and explicitly |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
183 listing them risks orders being dropped because they changed size. */ |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
184 MAX_ALIGNMENT * 3, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
185 MAX_ALIGNMENT * 5, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
186 MAX_ALIGNMENT * 6, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
187 MAX_ALIGNMENT * 7, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
188 MAX_ALIGNMENT * 9, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
189 MAX_ALIGNMENT * 10, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
190 MAX_ALIGNMENT * 11, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
191 MAX_ALIGNMENT * 12, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
192 MAX_ALIGNMENT * 13, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
193 MAX_ALIGNMENT * 14, |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
194 MAX_ALIGNMENT * 15, |
0 | 195 sizeof (struct tree_decl_non_common), |
196 sizeof (struct tree_field_decl), | |
197 sizeof (struct tree_parm_decl), | |
198 sizeof (struct tree_var_decl), | |
111 | 199 sizeof (struct tree_type_non_common), |
0 | 200 sizeof (struct function), |
201 sizeof (struct basic_block_def), | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
202 sizeof (struct cgraph_node), |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
203 sizeof (struct loop), |
0 | 204 }; |
205 | |
206 /* The total number of orders. */ | |
207 | |
208 #define NUM_ORDERS (HOST_BITS_PER_PTR + NUM_EXTRA_ORDERS) | |
209 | |
210 /* Compute the smallest nonnegative number which when added to X gives | |
211 a multiple of F. */ | |
212 | |
213 #define ROUND_UP_VALUE(x, f) ((f) - 1 - ((f) - 1 + (x)) % (f)) | |
214 | |
111 | 215 /* Round X to next multiple of the page size */ |
216 | |
217 #define PAGE_ALIGN(x) ROUND_UP ((x), G.pagesize) | |
0 | 218 |
219 /* The Ith entry is the number of objects on a page or order I. */ | |
220 | |
221 static unsigned objects_per_page_table[NUM_ORDERS]; | |
222 | |
223 /* The Ith entry is the size of an object on a page of order I. */ | |
224 | |
225 static size_t object_size_table[NUM_ORDERS]; | |
226 | |
227 /* The Ith entry is a pair of numbers (mult, shift) such that | |
228 ((k * mult) >> shift) mod 2^32 == (k / OBJECT_SIZE(I)) mod 2^32, | |
229 for all k evenly divisible by OBJECT_SIZE(I). */ | |
230 | |
231 static struct | |
232 { | |
233 size_t mult; | |
234 unsigned int shift; | |
235 } | |
236 inverse_table[NUM_ORDERS]; | |
237 | |
238 /* A page_entry records the status of an allocation page. This | |
239 structure is dynamically sized to fit the bitmap in_use_p. */ | |
111 | 240 struct page_entry |
0 | 241 { |
242 /* The next page-entry with objects of the same size, or NULL if | |
243 this is the last page-entry. */ | |
244 struct page_entry *next; | |
245 | |
246 /* The previous page-entry with objects of the same size, or NULL if | |
247 this is the first page-entry. The PREV pointer exists solely to | |
248 keep the cost of ggc_free manageable. */ | |
249 struct page_entry *prev; | |
250 | |
251 /* The number of bytes allocated. (This will always be a multiple | |
252 of the host system page size.) */ | |
253 size_t bytes; | |
254 | |
255 /* The address at which the memory is allocated. */ | |
256 char *page; | |
257 | |
258 #ifdef USING_MALLOC_PAGE_GROUPS | |
259 /* Back pointer to the page group this page came from. */ | |
260 struct page_group *group; | |
261 #endif | |
262 | |
263 /* This is the index in the by_depth varray where this page table | |
264 can be found. */ | |
265 unsigned long index_by_depth; | |
266 | |
267 /* Context depth of this page. */ | |
268 unsigned short context_depth; | |
269 | |
270 /* The number of free objects remaining on this page. */ | |
271 unsigned short num_free_objects; | |
272 | |
273 /* A likely candidate for the bit position of a free object for the | |
274 next allocation from this page. */ | |
275 unsigned short next_bit_hint; | |
276 | |
277 /* The lg of size of objects allocated from this page. */ | |
278 unsigned char order; | |
279 | |
111 | 280 /* Discarded page? */ |
281 bool discarded; | |
282 | |
0 | 283 /* A bit vector indicating whether or not objects are in use. The |
284 Nth bit is one if the Nth object on this page is allocated. This | |
285 array is dynamically sized. */ | |
286 unsigned long in_use_p[1]; | |
111 | 287 }; |
0 | 288 |
289 #ifdef USING_MALLOC_PAGE_GROUPS | |
290 /* A page_group describes a large allocation from malloc, from which | |
291 we parcel out aligned pages. */ | |
111 | 292 struct page_group |
0 | 293 { |
294 /* A linked list of all extant page groups. */ | |
295 struct page_group *next; | |
296 | |
297 /* The address we received from malloc. */ | |
298 char *allocation; | |
299 | |
300 /* The size of the block. */ | |
301 size_t alloc_size; | |
302 | |
303 /* A bitmask of pages in use. */ | |
304 unsigned int in_use; | |
111 | 305 }; |
0 | 306 #endif |
307 | |
308 #if HOST_BITS_PER_PTR <= 32 | |
309 | |
310 /* On 32-bit hosts, we use a two level page table, as pictured above. */ | |
311 typedef page_entry **page_table[PAGE_L1_SIZE]; | |
312 | |
313 #else | |
314 | |
315 /* On 64-bit hosts, we use the same two level page tables plus a linked | |
316 list that disambiguates the top 32-bits. There will almost always be | |
317 exactly one entry in the list. */ | |
318 typedef struct page_table_chain | |
319 { | |
320 struct page_table_chain *next; | |
321 size_t high_bits; | |
322 page_entry **table[PAGE_L1_SIZE]; | |
323 } *page_table; | |
324 | |
325 #endif | |
326 | |
111 | 327 class finalizer |
328 { | |
329 public: | |
330 finalizer (void *addr, void (*f)(void *)) : m_addr (addr), m_function (f) {} | |
331 | |
332 void *addr () const { return m_addr; } | |
333 | |
334 void call () const { m_function (m_addr); } | |
335 | |
336 private: | |
337 void *m_addr; | |
338 void (*m_function)(void *); | |
339 }; | |
340 | |
341 class vec_finalizer | |
342 { | |
343 public: | |
344 vec_finalizer (uintptr_t addr, void (*f)(void *), size_t s, size_t n) : | |
345 m_addr (addr), m_function (f), m_object_size (s), m_n_objects (n) {} | |
346 | |
347 void call () const | |
348 { | |
349 for (size_t i = 0; i < m_n_objects; i++) | |
350 m_function (reinterpret_cast<void *> (m_addr + (i * m_object_size))); | |
351 } | |
352 | |
353 void *addr () const { return reinterpret_cast<void *> (m_addr); } | |
354 | |
355 private: | |
356 uintptr_t m_addr; | |
357 void (*m_function)(void *); | |
358 size_t m_object_size; | |
359 size_t m_n_objects; | |
360 }; | |
361 | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
362 #ifdef ENABLE_GC_ALWAYS_COLLECT |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
363 /* List of free objects to be verified as actually free on the |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
364 next collection. */ |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
365 struct free_object |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
366 { |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
367 void *object; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
368 struct free_object *next; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
369 }; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
370 #endif |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
371 |
0 | 372 /* The rest of the global variables. */ |
111 | 373 static struct ggc_globals |
0 | 374 { |
375 /* The Nth element in this array is a page with objects of size 2^N. | |
376 If there are any pages with free objects, they will be at the | |
377 head of the list. NULL if there are no page-entries for this | |
378 object size. */ | |
379 page_entry *pages[NUM_ORDERS]; | |
380 | |
381 /* The Nth element in this array is the last page with objects of | |
382 size 2^N. NULL if there are no page-entries for this object | |
383 size. */ | |
384 page_entry *page_tails[NUM_ORDERS]; | |
385 | |
386 /* Lookup table for associating allocation pages with object addresses. */ | |
387 page_table lookup; | |
388 | |
389 /* The system's page size. */ | |
390 size_t pagesize; | |
391 size_t lg_pagesize; | |
392 | |
393 /* Bytes currently allocated. */ | |
394 size_t allocated; | |
395 | |
396 /* Bytes currently allocated at the end of the last collection. */ | |
397 size_t allocated_last_gc; | |
398 | |
399 /* Total amount of memory mapped. */ | |
400 size_t bytes_mapped; | |
401 | |
402 /* Bit N set if any allocations have been done at context depth N. */ | |
403 unsigned long context_depth_allocations; | |
404 | |
405 /* Bit N set if any collections have been done at context depth N. */ | |
406 unsigned long context_depth_collections; | |
407 | |
408 /* The current depth in the context stack. */ | |
409 unsigned short context_depth; | |
410 | |
411 /* A file descriptor open to /dev/zero for reading. */ | |
412 #if defined (HAVE_MMAP_DEV_ZERO) | |
413 int dev_zero_fd; | |
414 #endif | |
415 | |
416 /* A cache of free system pages. */ | |
417 page_entry *free_pages; | |
418 | |
419 #ifdef USING_MALLOC_PAGE_GROUPS | |
420 page_group *page_groups; | |
421 #endif | |
422 | |
423 /* The file descriptor for debugging output. */ | |
424 FILE *debug_file; | |
425 | |
426 /* Current number of elements in use in depth below. */ | |
427 unsigned int depth_in_use; | |
428 | |
429 /* Maximum number of elements that can be used before resizing. */ | |
430 unsigned int depth_max; | |
431 | |
432 /* Each element of this array is an index in by_depth where the given | |
433 depth starts. This structure is indexed by that given depth we | |
434 are interested in. */ | |
435 unsigned int *depth; | |
436 | |
437 /* Current number of elements in use in by_depth below. */ | |
438 unsigned int by_depth_in_use; | |
439 | |
440 /* Maximum number of elements that can be used before resizing. */ | |
441 unsigned int by_depth_max; | |
442 | |
443 /* Each element of this array is a pointer to a page_entry, all | |
444 page_entries can be found in here by increasing depth. | |
445 index_by_depth in the page_entry is the index into this data | |
446 structure where that page_entry can be found. This is used to | |
447 speed up finding all page_entries at a particular depth. */ | |
448 page_entry **by_depth; | |
449 | |
450 /* Each element is a pointer to the saved in_use_p bits, if any, | |
451 zero otherwise. We allocate them all together, to enable a | |
452 better runtime data access pattern. */ | |
453 unsigned long **save_in_use; | |
454 | |
111 | 455 /* Finalizers for single objects. The first index is collection_depth. */ |
456 vec<vec<finalizer> > finalizers; | |
457 | |
458 /* Finalizers for vectors of objects. */ | |
459 vec<vec<vec_finalizer> > vec_finalizers; | |
460 | |
0 | 461 #ifdef ENABLE_GC_ALWAYS_COLLECT |
462 /* List of free objects to be verified as actually free on the | |
463 next collection. */ | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
464 struct free_object *free_object_list; |
0 | 465 #endif |
466 | |
467 struct | |
468 { | |
67
f6334be47118
update gcc from gcc-4.6-20100522 to gcc-4.6-20110318
nobuyasu <dimolto@cr.ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
469 /* Total GC-allocated memory. */ |
0 | 470 unsigned long long total_allocated; |
67
f6334be47118
update gcc from gcc-4.6-20100522 to gcc-4.6-20110318
nobuyasu <dimolto@cr.ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
471 /* Total overhead for GC-allocated memory. */ |
0 | 472 unsigned long long total_overhead; |
473 | |
474 /* Total allocations and overhead for sizes less than 32, 64 and 128. | |
475 These sizes are interesting because they are typical cache line | |
476 sizes. */ | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
477 |
0 | 478 unsigned long long total_allocated_under32; |
479 unsigned long long total_overhead_under32; | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
480 |
0 | 481 unsigned long long total_allocated_under64; |
482 unsigned long long total_overhead_under64; | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
483 |
0 | 484 unsigned long long total_allocated_under128; |
485 unsigned long long total_overhead_under128; | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
486 |
0 | 487 /* The allocations for each of the allocation orders. */ |
488 unsigned long long total_allocated_per_order[NUM_ORDERS]; | |
489 | |
490 /* The overhead for each of the allocation orders. */ | |
491 unsigned long long total_overhead_per_order[NUM_ORDERS]; | |
492 } stats; | |
493 } G; | |
494 | |
111 | 495 /* True if a gc is currently taking place. */ |
496 | |
497 static bool in_gc = false; | |
498 | |
0 | 499 /* The size in bytes required to maintain a bitmap for the objects |
500 on a page-entry. */ | |
501 #define BITMAP_SIZE(Num_objects) \ | |
111 | 502 (CEIL ((Num_objects), HOST_BITS_PER_LONG) * sizeof (long)) |
0 | 503 |
504 /* Allocate pages in chunks of this size, to throttle calls to memory | |
505 allocation routines. The first page is used, the rest go onto the | |
506 free list. This cannot be larger than HOST_BITS_PER_INT for the | |
507 in_use bitmask for page_group. Hosts that need a different value | |
508 can override this by defining GGC_QUIRE_SIZE explicitly. */ | |
509 #ifndef GGC_QUIRE_SIZE | |
510 # ifdef USING_MMAP | |
111 | 511 # define GGC_QUIRE_SIZE 512 /* 2MB for 4K pages */ |
0 | 512 # else |
513 # define GGC_QUIRE_SIZE 16 | |
514 # endif | |
515 #endif | |
516 | |
517 /* Initial guess as to how many page table entries we might need. */ | |
518 #define INITIAL_PTE_COUNT 128 | |
519 | |
520 static page_entry *lookup_page_table_entry (const void *); | |
521 static void set_page_table_entry (void *, page_entry *); | |
522 #ifdef USING_MMAP | |
111 | 523 static char *alloc_anon (char *, size_t, bool check); |
0 | 524 #endif |
525 #ifdef USING_MALLOC_PAGE_GROUPS | |
526 static size_t page_group_index (char *, char *); | |
527 static void set_page_group_in_use (page_group *, char *); | |
528 static void clear_page_group_in_use (page_group *, char *); | |
529 #endif | |
530 static struct page_entry * alloc_page (unsigned); | |
531 static void free_page (struct page_entry *); | |
532 static void release_pages (void); | |
533 static void clear_marks (void); | |
534 static void sweep_pages (void); | |
535 static void ggc_recalculate_in_use_p (page_entry *); | |
536 static void compute_inverse (unsigned); | |
537 static inline void adjust_depth (void); | |
538 static void move_ptes_to_front (int, int); | |
539 | |
540 void debug_print_page_list (int); | |
541 static void push_depth (unsigned int); | |
542 static void push_by_depth (page_entry *, unsigned long *); | |
543 | |
544 /* Push an entry onto G.depth. */ | |
545 | |
546 inline static void | |
547 push_depth (unsigned int i) | |
548 { | |
549 if (G.depth_in_use >= G.depth_max) | |
550 { | |
551 G.depth_max *= 2; | |
552 G.depth = XRESIZEVEC (unsigned int, G.depth, G.depth_max); | |
553 } | |
554 G.depth[G.depth_in_use++] = i; | |
555 } | |
556 | |
557 /* Push an entry onto G.by_depth and G.save_in_use. */ | |
558 | |
559 inline static void | |
560 push_by_depth (page_entry *p, unsigned long *s) | |
561 { | |
562 if (G.by_depth_in_use >= G.by_depth_max) | |
563 { | |
564 G.by_depth_max *= 2; | |
565 G.by_depth = XRESIZEVEC (page_entry *, G.by_depth, G.by_depth_max); | |
566 G.save_in_use = XRESIZEVEC (unsigned long *, G.save_in_use, | |
567 G.by_depth_max); | |
568 } | |
569 G.by_depth[G.by_depth_in_use] = p; | |
570 G.save_in_use[G.by_depth_in_use++] = s; | |
571 } | |
572 | |
573 #if (GCC_VERSION < 3001) | |
574 #define prefetch(X) ((void) X) | |
575 #else | |
576 #define prefetch(X) __builtin_prefetch (X) | |
577 #endif | |
578 | |
579 #define save_in_use_p_i(__i) \ | |
580 (G.save_in_use[__i]) | |
581 #define save_in_use_p(__p) \ | |
582 (save_in_use_p_i (__p->index_by_depth)) | |
583 | |
111 | 584 /* Traverse the page table and find the entry for a page. |
585 If the object wasn't allocated in GC return NULL. */ | |
586 | |
587 static inline page_entry * | |
588 safe_lookup_page_table_entry (const void *p) | |
0 | 589 { |
590 page_entry ***base; | |
591 size_t L1, L2; | |
592 | |
593 #if HOST_BITS_PER_PTR <= 32 | |
594 base = &G.lookup[0]; | |
595 #else | |
596 page_table table = G.lookup; | |
111 | 597 uintptr_t high_bits = (uintptr_t) p & ~ (uintptr_t) 0xffffffff; |
0 | 598 while (1) |
599 { | |
600 if (table == NULL) | |
111 | 601 return NULL; |
0 | 602 if (table->high_bits == high_bits) |
603 break; | |
604 table = table->next; | |
605 } | |
606 base = &table->table[0]; | |
607 #endif | |
608 | |
609 /* Extract the level 1 and 2 indices. */ | |
610 L1 = LOOKUP_L1 (p); | |
611 L2 = LOOKUP_L2 (p); | |
111 | 612 if (! base[L1]) |
613 return NULL; | |
614 | |
615 return base[L1][L2]; | |
0 | 616 } |
617 | |
618 /* Traverse the page table and find the entry for a page. | |
619 Die (probably) if the object wasn't allocated via GC. */ | |
620 | |
621 static inline page_entry * | |
622 lookup_page_table_entry (const void *p) | |
623 { | |
624 page_entry ***base; | |
625 size_t L1, L2; | |
626 | |
627 #if HOST_BITS_PER_PTR <= 32 | |
628 base = &G.lookup[0]; | |
629 #else | |
630 page_table table = G.lookup; | |
111 | 631 uintptr_t high_bits = (uintptr_t) p & ~ (uintptr_t) 0xffffffff; |
0 | 632 while (table->high_bits != high_bits) |
633 table = table->next; | |
634 base = &table->table[0]; | |
635 #endif | |
636 | |
637 /* Extract the level 1 and 2 indices. */ | |
638 L1 = LOOKUP_L1 (p); | |
639 L2 = LOOKUP_L2 (p); | |
640 | |
641 return base[L1][L2]; | |
642 } | |
643 | |
644 /* Set the page table entry for a page. */ | |
645 | |
646 static void | |
647 set_page_table_entry (void *p, page_entry *entry) | |
648 { | |
649 page_entry ***base; | |
650 size_t L1, L2; | |
651 | |
652 #if HOST_BITS_PER_PTR <= 32 | |
653 base = &G.lookup[0]; | |
654 #else | |
655 page_table table; | |
111 | 656 uintptr_t high_bits = (uintptr_t) p & ~ (uintptr_t) 0xffffffff; |
0 | 657 for (table = G.lookup; table; table = table->next) |
658 if (table->high_bits == high_bits) | |
659 goto found; | |
660 | |
661 /* Not found -- allocate a new table. */ | |
662 table = XCNEW (struct page_table_chain); | |
663 table->next = G.lookup; | |
664 table->high_bits = high_bits; | |
665 G.lookup = table; | |
666 found: | |
667 base = &table->table[0]; | |
668 #endif | |
669 | |
670 /* Extract the level 1 and 2 indices. */ | |
671 L1 = LOOKUP_L1 (p); | |
672 L2 = LOOKUP_L2 (p); | |
673 | |
674 if (base[L1] == NULL) | |
675 base[L1] = XCNEWVEC (page_entry *, PAGE_L2_SIZE); | |
676 | |
677 base[L1][L2] = entry; | |
678 } | |
679 | |
680 /* Prints the page-entry for object size ORDER, for debugging. */ | |
681 | |
67
f6334be47118
update gcc from gcc-4.6-20100522 to gcc-4.6-20110318
nobuyasu <dimolto@cr.ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
682 DEBUG_FUNCTION void |
0 | 683 debug_print_page_list (int order) |
684 { | |
685 page_entry *p; | |
686 printf ("Head=%p, Tail=%p:\n", (void *) G.pages[order], | |
687 (void *) G.page_tails[order]); | |
688 p = G.pages[order]; | |
689 while (p != NULL) | |
690 { | |
691 printf ("%p(%1d|%3d) -> ", (void *) p, p->context_depth, | |
692 p->num_free_objects); | |
693 p = p->next; | |
694 } | |
695 printf ("NULL\n"); | |
696 fflush (stdout); | |
697 } | |
698 | |
699 #ifdef USING_MMAP | |
700 /* Allocate SIZE bytes of anonymous memory, preferably near PREF, | |
701 (if non-null). The ifdef structure here is intended to cause a | |
702 compile error unless exactly one of the HAVE_* is defined. */ | |
703 | |
704 static inline char * | |
111 | 705 alloc_anon (char *pref ATTRIBUTE_UNUSED, size_t size, bool check) |
0 | 706 { |
707 #ifdef HAVE_MMAP_ANON | |
708 char *page = (char *) mmap (pref, size, PROT_READ | PROT_WRITE, | |
709 MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); | |
710 #endif | |
711 #ifdef HAVE_MMAP_DEV_ZERO | |
712 char *page = (char *) mmap (pref, size, PROT_READ | PROT_WRITE, | |
713 MAP_PRIVATE, G.dev_zero_fd, 0); | |
714 #endif | |
715 | |
716 if (page == (char *) MAP_FAILED) | |
717 { | |
111 | 718 if (!check) |
719 return NULL; | |
0 | 720 perror ("virtual memory exhausted"); |
721 exit (FATAL_EXIT_CODE); | |
722 } | |
723 | |
724 /* Remember that we allocated this memory. */ | |
725 G.bytes_mapped += size; | |
726 | |
727 /* Pretend we don't have access to the allocated pages. We'll enable | |
67
f6334be47118
update gcc from gcc-4.6-20100522 to gcc-4.6-20110318
nobuyasu <dimolto@cr.ie.u-ryukyu.ac.jp>
parents:
55
diff
changeset
|
728 access to smaller pieces of the area in ggc_internal_alloc. Discard the |
0 | 729 handle to avoid handle leak. */ |
730 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (page, size)); | |
731 | |
732 return page; | |
733 } | |
734 #endif | |
735 #ifdef USING_MALLOC_PAGE_GROUPS | |
736 /* Compute the index for this page into the page group. */ | |
737 | |
738 static inline size_t | |
739 page_group_index (char *allocation, char *page) | |
740 { | |
741 return (size_t) (page - allocation) >> G.lg_pagesize; | |
742 } | |
743 | |
744 /* Set and clear the in_use bit for this page in the page group. */ | |
745 | |
746 static inline void | |
747 set_page_group_in_use (page_group *group, char *page) | |
748 { | |
749 group->in_use |= 1 << page_group_index (group->allocation, page); | |
750 } | |
751 | |
752 static inline void | |
753 clear_page_group_in_use (page_group *group, char *page) | |
754 { | |
755 group->in_use &= ~(1 << page_group_index (group->allocation, page)); | |
756 } | |
757 #endif | |
758 | |
759 /* Allocate a new page for allocating objects of size 2^ORDER, | |
760 and return an entry for it. The entry is not added to the | |
761 appropriate page_table list. */ | |
762 | |
763 static inline struct page_entry * | |
764 alloc_page (unsigned order) | |
765 { | |
766 struct page_entry *entry, *p, **pp; | |
767 char *page; | |
768 size_t num_objects; | |
769 size_t bitmap_size; | |
770 size_t page_entry_size; | |
771 size_t entry_size; | |
772 #ifdef USING_MALLOC_PAGE_GROUPS | |
773 page_group *group; | |
774 #endif | |
775 | |
776 num_objects = OBJECTS_PER_PAGE (order); | |
777 bitmap_size = BITMAP_SIZE (num_objects + 1); | |
778 page_entry_size = sizeof (page_entry) - sizeof (long) + bitmap_size; | |
779 entry_size = num_objects * OBJECT_SIZE (order); | |
780 if (entry_size < G.pagesize) | |
781 entry_size = G.pagesize; | |
111 | 782 entry_size = PAGE_ALIGN (entry_size); |
0 | 783 |
784 entry = NULL; | |
785 page = NULL; | |
786 | |
787 /* Check the list of free pages for one we can use. */ | |
788 for (pp = &G.free_pages, p = *pp; p; pp = &p->next, p = *pp) | |
789 if (p->bytes == entry_size) | |
790 break; | |
791 | |
792 if (p != NULL) | |
793 { | |
111 | 794 if (p->discarded) |
795 G.bytes_mapped += p->bytes; | |
796 p->discarded = false; | |
797 | |
0 | 798 /* Recycle the allocated memory from this page ... */ |
799 *pp = p->next; | |
800 page = p->page; | |
801 | |
802 #ifdef USING_MALLOC_PAGE_GROUPS | |
803 group = p->group; | |
804 #endif | |
805 | |
806 /* ... and, if possible, the page entry itself. */ | |
807 if (p->order == order) | |
808 { | |
809 entry = p; | |
810 memset (entry, 0, page_entry_size); | |
811 } | |
812 else | |
813 free (p); | |
814 } | |
815 #ifdef USING_MMAP | |
816 else if (entry_size == G.pagesize) | |
817 { | |
818 /* We want just one page. Allocate a bunch of them and put the | |
819 extras on the freelist. (Can only do this optimization with | |
820 mmap for backing store.) */ | |
821 struct page_entry *e, *f = G.free_pages; | |
111 | 822 int i, entries = GGC_QUIRE_SIZE; |
823 | |
824 page = alloc_anon (NULL, G.pagesize * GGC_QUIRE_SIZE, false); | |
825 if (page == NULL) | |
826 { | |
827 page = alloc_anon (NULL, G.pagesize, true); | |
828 entries = 1; | |
829 } | |
0 | 830 |
831 /* This loop counts down so that the chain will be in ascending | |
832 memory order. */ | |
111 | 833 for (i = entries - 1; i >= 1; i--) |
0 | 834 { |
835 e = XCNEWVAR (struct page_entry, page_entry_size); | |
836 e->order = order; | |
837 e->bytes = G.pagesize; | |
838 e->page = page + (i << G.lg_pagesize); | |
839 e->next = f; | |
840 f = e; | |
841 } | |
842 | |
843 G.free_pages = f; | |
844 } | |
845 else | |
111 | 846 page = alloc_anon (NULL, entry_size, true); |
0 | 847 #endif |
848 #ifdef USING_MALLOC_PAGE_GROUPS | |
849 else | |
850 { | |
851 /* Allocate a large block of memory and serve out the aligned | |
852 pages therein. This results in much less memory wastage | |
853 than the traditional implementation of valloc. */ | |
854 | |
855 char *allocation, *a, *enda; | |
856 size_t alloc_size, head_slop, tail_slop; | |
857 int multiple_pages = (entry_size == G.pagesize); | |
858 | |
859 if (multiple_pages) | |
860 alloc_size = GGC_QUIRE_SIZE * G.pagesize; | |
861 else | |
862 alloc_size = entry_size + G.pagesize - 1; | |
863 allocation = XNEWVEC (char, alloc_size); | |
864 | |
111 | 865 page = (char *) (((uintptr_t) allocation + G.pagesize - 1) & -G.pagesize); |
0 | 866 head_slop = page - allocation; |
867 if (multiple_pages) | |
868 tail_slop = ((size_t) allocation + alloc_size) & (G.pagesize - 1); | |
869 else | |
870 tail_slop = alloc_size - entry_size - head_slop; | |
871 enda = allocation + alloc_size - tail_slop; | |
872 | |
873 /* We allocated N pages, which are likely not aligned, leaving | |
874 us with N-1 usable pages. We plan to place the page_group | |
875 structure somewhere in the slop. */ | |
876 if (head_slop >= sizeof (page_group)) | |
877 group = (page_group *)page - 1; | |
878 else | |
879 { | |
880 /* We magically got an aligned allocation. Too bad, we have | |
881 to waste a page anyway. */ | |
882 if (tail_slop == 0) | |
883 { | |
884 enda -= G.pagesize; | |
885 tail_slop += G.pagesize; | |
886 } | |
887 gcc_assert (tail_slop >= sizeof (page_group)); | |
888 group = (page_group *)enda; | |
889 tail_slop -= sizeof (page_group); | |
890 } | |
891 | |
892 /* Remember that we allocated this memory. */ | |
893 group->next = G.page_groups; | |
894 group->allocation = allocation; | |
895 group->alloc_size = alloc_size; | |
896 group->in_use = 0; | |
897 G.page_groups = group; | |
898 G.bytes_mapped += alloc_size; | |
899 | |
900 /* If we allocated multiple pages, put the rest on the free list. */ | |
901 if (multiple_pages) | |
902 { | |
903 struct page_entry *e, *f = G.free_pages; | |
904 for (a = enda - G.pagesize; a != page; a -= G.pagesize) | |
905 { | |
906 e = XCNEWVAR (struct page_entry, page_entry_size); | |
907 e->order = order; | |
908 e->bytes = G.pagesize; | |
909 e->page = a; | |
910 e->group = group; | |
911 e->next = f; | |
912 f = e; | |
913 } | |
914 G.free_pages = f; | |
915 } | |
916 } | |
917 #endif | |
918 | |
919 if (entry == NULL) | |
920 entry = XCNEWVAR (struct page_entry, page_entry_size); | |
921 | |
922 entry->bytes = entry_size; | |
923 entry->page = page; | |
924 entry->context_depth = G.context_depth; | |
925 entry->order = order; | |
926 entry->num_free_objects = num_objects; | |
927 entry->next_bit_hint = 1; | |
928 | |
929 G.context_depth_allocations |= (unsigned long)1 << G.context_depth; | |
930 | |
931 #ifdef USING_MALLOC_PAGE_GROUPS | |
932 entry->group = group; | |
933 set_page_group_in_use (group, page); | |
934 #endif | |
935 | |
936 /* Set the one-past-the-end in-use bit. This acts as a sentry as we | |
937 increment the hint. */ | |
938 entry->in_use_p[num_objects / HOST_BITS_PER_LONG] | |
939 = (unsigned long) 1 << (num_objects % HOST_BITS_PER_LONG); | |
940 | |
941 set_page_table_entry (page, entry); | |
942 | |
943 if (GGC_DEBUG_LEVEL >= 2) | |
944 fprintf (G.debug_file, | |
945 "Allocating page at %p, object size=%lu, data %p-%p\n", | |
946 (void *) entry, (unsigned long) OBJECT_SIZE (order), page, | |
947 page + entry_size - 1); | |
948 | |
949 return entry; | |
950 } | |
951 | |
952 /* Adjust the size of G.depth so that no index greater than the one | |
953 used by the top of the G.by_depth is used. */ | |
954 | |
955 static inline void | |
956 adjust_depth (void) | |
957 { | |
958 page_entry *top; | |
959 | |
960 if (G.by_depth_in_use) | |
961 { | |
962 top = G.by_depth[G.by_depth_in_use-1]; | |
963 | |
964 /* Peel back indices in depth that index into by_depth, so that | |
965 as new elements are added to by_depth, we note the indices | |
966 of those elements, if they are for new context depths. */ | |
967 while (G.depth_in_use > (size_t)top->context_depth+1) | |
968 --G.depth_in_use; | |
969 } | |
970 } | |
971 | |
972 /* For a page that is no longer needed, put it on the free page list. */ | |
973 | |
974 static void | |
975 free_page (page_entry *entry) | |
976 { | |
977 if (GGC_DEBUG_LEVEL >= 2) | |
978 fprintf (G.debug_file, | |
979 "Deallocating page at %p, data %p-%p\n", (void *) entry, | |
980 entry->page, entry->page + entry->bytes - 1); | |
981 | |
982 /* Mark the page as inaccessible. Discard the handle to avoid handle | |
983 leak. */ | |
984 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (entry->page, entry->bytes)); | |
985 | |
986 set_page_table_entry (entry->page, NULL); | |
987 | |
988 #ifdef USING_MALLOC_PAGE_GROUPS | |
989 clear_page_group_in_use (entry->group, entry->page); | |
990 #endif | |
991 | |
992 if (G.by_depth_in_use > 1) | |
993 { | |
994 page_entry *top = G.by_depth[G.by_depth_in_use-1]; | |
995 int i = entry->index_by_depth; | |
996 | |
997 /* We cannot free a page from a context deeper than the current | |
998 one. */ | |
999 gcc_assert (entry->context_depth == top->context_depth); | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1000 |
0 | 1001 /* Put top element into freed slot. */ |
1002 G.by_depth[i] = top; | |
1003 G.save_in_use[i] = G.save_in_use[G.by_depth_in_use-1]; | |
1004 top->index_by_depth = i; | |
1005 } | |
1006 --G.by_depth_in_use; | |
1007 | |
1008 adjust_depth (); | |
1009 | |
1010 entry->next = G.free_pages; | |
1011 G.free_pages = entry; | |
1012 } | |
1013 | |
1014 /* Release the free page cache to the system. */ | |
1015 | |
1016 static void | |
1017 release_pages (void) | |
1018 { | |
111 | 1019 #ifdef USING_MADVISE |
1020 page_entry *p, *start_p; | |
1021 char *start; | |
1022 size_t len; | |
1023 size_t mapped_len; | |
1024 page_entry *next, *prev, *newprev; | |
1025 size_t free_unit = (GGC_QUIRE_SIZE/2) * G.pagesize; | |
1026 | |
1027 /* First free larger continuous areas to the OS. | |
1028 This allows other allocators to grab these areas if needed. | |
1029 This is only done on larger chunks to avoid fragmentation. | |
1030 This does not always work because the free_pages list is only | |
1031 approximately sorted. */ | |
1032 | |
1033 p = G.free_pages; | |
1034 prev = NULL; | |
1035 while (p) | |
1036 { | |
1037 start = p->page; | |
1038 start_p = p; | |
1039 len = 0; | |
1040 mapped_len = 0; | |
1041 newprev = prev; | |
1042 while (p && p->page == start + len) | |
1043 { | |
1044 len += p->bytes; | |
1045 if (!p->discarded) | |
1046 mapped_len += p->bytes; | |
1047 newprev = p; | |
1048 p = p->next; | |
1049 } | |
1050 if (len >= free_unit) | |
1051 { | |
1052 while (start_p != p) | |
1053 { | |
1054 next = start_p->next; | |
1055 free (start_p); | |
1056 start_p = next; | |
1057 } | |
1058 munmap (start, len); | |
1059 if (prev) | |
1060 prev->next = p; | |
1061 else | |
1062 G.free_pages = p; | |
1063 G.bytes_mapped -= mapped_len; | |
1064 continue; | |
1065 } | |
1066 prev = newprev; | |
1067 } | |
1068 | |
1069 /* Now give back the fragmented pages to the OS, but keep the address | |
1070 space to reuse it next time. */ | |
1071 | |
1072 for (p = G.free_pages; p; ) | |
1073 { | |
1074 if (p->discarded) | |
1075 { | |
1076 p = p->next; | |
1077 continue; | |
1078 } | |
1079 start = p->page; | |
1080 len = p->bytes; | |
1081 start_p = p; | |
1082 p = p->next; | |
1083 while (p && p->page == start + len) | |
1084 { | |
1085 len += p->bytes; | |
1086 p = p->next; | |
1087 } | |
1088 /* Give the page back to the kernel, but don't free the mapping. | |
1089 This avoids fragmentation in the virtual memory map of the | |
1090 process. Next time we can reuse it by just touching it. */ | |
1091 madvise (start, len, MADV_DONTNEED); | |
1092 /* Don't count those pages as mapped to not touch the garbage collector | |
1093 unnecessarily. */ | |
1094 G.bytes_mapped -= len; | |
1095 while (start_p != p) | |
1096 { | |
1097 start_p->discarded = true; | |
1098 start_p = start_p->next; | |
1099 } | |
1100 } | |
1101 #endif | |
1102 #if defined(USING_MMAP) && !defined(USING_MADVISE) | |
0 | 1103 page_entry *p, *next; |
1104 char *start; | |
1105 size_t len; | |
1106 | |
1107 /* Gather up adjacent pages so they are unmapped together. */ | |
1108 p = G.free_pages; | |
1109 | |
1110 while (p) | |
1111 { | |
1112 start = p->page; | |
1113 next = p->next; | |
1114 len = p->bytes; | |
1115 free (p); | |
1116 p = next; | |
1117 | |
1118 while (p && p->page == start + len) | |
1119 { | |
1120 next = p->next; | |
1121 len += p->bytes; | |
1122 free (p); | |
1123 p = next; | |
1124 } | |
1125 | |
1126 munmap (start, len); | |
1127 G.bytes_mapped -= len; | |
1128 } | |
1129 | |
1130 G.free_pages = NULL; | |
1131 #endif | |
1132 #ifdef USING_MALLOC_PAGE_GROUPS | |
1133 page_entry **pp, *p; | |
1134 page_group **gp, *g; | |
1135 | |
1136 /* Remove all pages from free page groups from the list. */ | |
1137 pp = &G.free_pages; | |
1138 while ((p = *pp) != NULL) | |
1139 if (p->group->in_use == 0) | |
1140 { | |
1141 *pp = p->next; | |
1142 free (p); | |
1143 } | |
1144 else | |
1145 pp = &p->next; | |
1146 | |
1147 /* Remove all free page groups, and release the storage. */ | |
1148 gp = &G.page_groups; | |
1149 while ((g = *gp) != NULL) | |
1150 if (g->in_use == 0) | |
1151 { | |
1152 *gp = g->next; | |
1153 G.bytes_mapped -= g->alloc_size; | |
1154 free (g->allocation); | |
1155 } | |
1156 else | |
1157 gp = &g->next; | |
1158 #endif | |
1159 } | |
1160 | |
1161 /* This table provides a fast way to determine ceil(log_2(size)) for | |
1162 allocation requests. The minimum allocation size is eight bytes. */ | |
1163 #define NUM_SIZE_LOOKUP 512 | |
1164 static unsigned char size_lookup[NUM_SIZE_LOOKUP] = | |
1165 { | |
1166 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, | |
1167 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, | |
1168 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, | |
1169 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, | |
1170 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
1171 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
1172 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
1173 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
1174 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1175 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1176 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1177 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1178 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1179 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1180 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1181 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
1182 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1183 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1184 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1185 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1186 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1187 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1188 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1189 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1190 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1191 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1192 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1193 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1194 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1195 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1196 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
1197 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9 | |
1198 }; | |
1199 | |
111 | 1200 /* For a given size of memory requested for allocation, return the |
1201 actual size that is going to be allocated, as well as the size | |
1202 order. */ | |
1203 | |
1204 static void | |
1205 ggc_round_alloc_size_1 (size_t requested_size, | |
1206 size_t *size_order, | |
1207 size_t *alloced_size) | |
0 | 1208 { |
111 | 1209 size_t order, object_size; |
1210 | |
1211 if (requested_size < NUM_SIZE_LOOKUP) | |
1212 { | |
1213 order = size_lookup[requested_size]; | |
1214 object_size = OBJECT_SIZE (order); | |
1215 } | |
1216 else | |
1217 { | |
1218 order = 10; | |
1219 while (requested_size > (object_size = OBJECT_SIZE (order))) | |
1220 order++; | |
1221 } | |
1222 | |
1223 if (size_order) | |
1224 *size_order = order; | |
1225 if (alloced_size) | |
1226 *alloced_size = object_size; | |
1227 } | |
1228 | |
1229 /* For a given size of memory requested for allocation, return the | |
1230 actual size that is going to be allocated. */ | |
1231 | |
1232 size_t | |
1233 ggc_round_alloc_size (size_t requested_size) | |
1234 { | |
1235 size_t size = 0; | |
1236 | |
1237 ggc_round_alloc_size_1 (requested_size, NULL, &size); | |
1238 return size; | |
1239 } | |
1240 | |
1241 /* Push a finalizer onto the appropriate vec. */ | |
1242 | |
1243 static void | |
1244 add_finalizer (void *result, void (*f)(void *), size_t s, size_t n) | |
1245 { | |
1246 if (f == NULL) | |
1247 /* No finalizer. */; | |
1248 else if (n == 1) | |
1249 { | |
1250 finalizer fin (result, f); | |
1251 G.finalizers[G.context_depth].safe_push (fin); | |
1252 } | |
1253 else | |
1254 { | |
1255 vec_finalizer fin (reinterpret_cast<uintptr_t> (result), f, s, n); | |
1256 G.vec_finalizers[G.context_depth].safe_push (fin); | |
1257 } | |
0 | 1258 } |
1259 | |
1260 /* Allocate a chunk of memory of SIZE bytes. Its contents are undefined. */ | |
1261 | |
1262 void * | |
111 | 1263 ggc_internal_alloc (size_t size, void (*f)(void *), size_t s, size_t n |
1264 MEM_STAT_DECL) | |
0 | 1265 { |
1266 size_t order, word, bit, object_offset, object_size; | |
1267 struct page_entry *entry; | |
1268 void *result; | |
1269 | |
111 | 1270 ggc_round_alloc_size_1 (size, &order, &object_size); |
0 | 1271 |
1272 /* If there are non-full pages for this size allocation, they are at | |
1273 the head of the list. */ | |
1274 entry = G.pages[order]; | |
1275 | |
1276 /* If there is no page for this object size, or all pages in this | |
1277 context are full, allocate a new page. */ | |
1278 if (entry == NULL || entry->num_free_objects == 0) | |
1279 { | |
1280 struct page_entry *new_entry; | |
1281 new_entry = alloc_page (order); | |
1282 | |
1283 new_entry->index_by_depth = G.by_depth_in_use; | |
1284 push_by_depth (new_entry, 0); | |
1285 | |
1286 /* We can skip context depths, if we do, make sure we go all the | |
1287 way to the new depth. */ | |
1288 while (new_entry->context_depth >= G.depth_in_use) | |
1289 push_depth (G.by_depth_in_use-1); | |
1290 | |
1291 /* If this is the only entry, it's also the tail. If it is not | |
1292 the only entry, then we must update the PREV pointer of the | |
1293 ENTRY (G.pages[order]) to point to our new page entry. */ | |
1294 if (entry == NULL) | |
1295 G.page_tails[order] = new_entry; | |
1296 else | |
1297 entry->prev = new_entry; | |
1298 | |
1299 /* Put new pages at the head of the page list. By definition the | |
1300 entry at the head of the list always has a NULL pointer. */ | |
1301 new_entry->next = entry; | |
1302 new_entry->prev = NULL; | |
1303 entry = new_entry; | |
1304 G.pages[order] = new_entry; | |
1305 | |
1306 /* For a new page, we know the word and bit positions (in the | |
1307 in_use bitmap) of the first available object -- they're zero. */ | |
1308 new_entry->next_bit_hint = 1; | |
1309 word = 0; | |
1310 bit = 0; | |
1311 object_offset = 0; | |
1312 } | |
1313 else | |
1314 { | |
1315 /* First try to use the hint left from the previous allocation | |
1316 to locate a clear bit in the in-use bitmap. We've made sure | |
1317 that the one-past-the-end bit is always set, so if the hint | |
1318 has run over, this test will fail. */ | |
1319 unsigned hint = entry->next_bit_hint; | |
1320 word = hint / HOST_BITS_PER_LONG; | |
1321 bit = hint % HOST_BITS_PER_LONG; | |
1322 | |
1323 /* If the hint didn't work, scan the bitmap from the beginning. */ | |
1324 if ((entry->in_use_p[word] >> bit) & 1) | |
1325 { | |
1326 word = bit = 0; | |
1327 while (~entry->in_use_p[word] == 0) | |
1328 ++word; | |
1329 | |
1330 #if GCC_VERSION >= 3004 | |
1331 bit = __builtin_ctzl (~entry->in_use_p[word]); | |
1332 #else | |
1333 while ((entry->in_use_p[word] >> bit) & 1) | |
1334 ++bit; | |
1335 #endif | |
1336 | |
1337 hint = word * HOST_BITS_PER_LONG + bit; | |
1338 } | |
1339 | |
1340 /* Next time, try the next bit. */ | |
1341 entry->next_bit_hint = hint + 1; | |
1342 | |
1343 object_offset = hint * object_size; | |
1344 } | |
1345 | |
1346 /* Set the in-use bit. */ | |
1347 entry->in_use_p[word] |= ((unsigned long) 1 << bit); | |
1348 | |
1349 /* Keep a running total of the number of free objects. If this page | |
1350 fills up, we may have to move it to the end of the list if the | |
1351 next page isn't full. If the next page is full, all subsequent | |
1352 pages are full, so there's no need to move it. */ | |
1353 if (--entry->num_free_objects == 0 | |
1354 && entry->next != NULL | |
1355 && entry->next->num_free_objects > 0) | |
1356 { | |
1357 /* We have a new head for the list. */ | |
1358 G.pages[order] = entry->next; | |
1359 | |
1360 /* We are moving ENTRY to the end of the page table list. | |
1361 The new page at the head of the list will have NULL in | |
1362 its PREV field and ENTRY will have NULL in its NEXT field. */ | |
1363 entry->next->prev = NULL; | |
1364 entry->next = NULL; | |
1365 | |
1366 /* Append ENTRY to the tail of the list. */ | |
1367 entry->prev = G.page_tails[order]; | |
1368 G.page_tails[order]->next = entry; | |
1369 G.page_tails[order] = entry; | |
1370 } | |
1371 | |
1372 /* Calculate the object's address. */ | |
1373 result = entry->page + object_offset; | |
111 | 1374 if (GATHER_STATISTICS) |
1375 ggc_record_overhead (OBJECT_SIZE (order), OBJECT_SIZE (order) - size, | |
1376 result FINAL_PASS_MEM_STAT); | |
0 | 1377 |
1378 #ifdef ENABLE_GC_CHECKING | |
1379 /* Keep poisoning-by-writing-0xaf the object, in an attempt to keep the | |
1380 exact same semantics in presence of memory bugs, regardless of | |
1381 ENABLE_VALGRIND_CHECKING. We override this request below. Drop the | |
1382 handle to avoid handle leak. */ | |
1383 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (result, object_size)); | |
1384 | |
1385 /* `Poison' the entire allocated object, including any padding at | |
1386 the end. */ | |
1387 memset (result, 0xaf, object_size); | |
1388 | |
1389 /* Make the bytes after the end of the object unaccessible. Discard the | |
1390 handle to avoid handle leak. */ | |
1391 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS ((char *) result + size, | |
1392 object_size - size)); | |
1393 #endif | |
1394 | |
1395 /* Tell Valgrind that the memory is there, but its content isn't | |
1396 defined. The bytes at the end of the object are still marked | |
1397 unaccessible. */ | |
1398 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (result, size)); | |
1399 | |
1400 /* Keep track of how many bytes are being allocated. This | |
1401 information is used in deciding when to collect. */ | |
1402 G.allocated += object_size; | |
1403 | |
1404 /* For timevar statistics. */ | |
1405 timevar_ggc_mem_total += object_size; | |
1406 | |
111 | 1407 if (f) |
1408 add_finalizer (result, f, s, n); | |
1409 | |
1410 if (GATHER_STATISTICS) | |
1411 { | |
1412 size_t overhead = object_size - size; | |
1413 | |
1414 G.stats.total_overhead += overhead; | |
1415 G.stats.total_allocated += object_size; | |
1416 G.stats.total_overhead_per_order[order] += overhead; | |
1417 G.stats.total_allocated_per_order[order] += object_size; | |
1418 | |
1419 if (size <= 32) | |
1420 { | |
1421 G.stats.total_overhead_under32 += overhead; | |
1422 G.stats.total_allocated_under32 += object_size; | |
1423 } | |
1424 if (size <= 64) | |
1425 { | |
1426 G.stats.total_overhead_under64 += overhead; | |
1427 G.stats.total_allocated_under64 += object_size; | |
1428 } | |
1429 if (size <= 128) | |
1430 { | |
1431 G.stats.total_overhead_under128 += overhead; | |
1432 G.stats.total_allocated_under128 += object_size; | |
1433 } | |
1434 } | |
0 | 1435 |
1436 if (GGC_DEBUG_LEVEL >= 3) | |
1437 fprintf (G.debug_file, | |
1438 "Allocating object, requested size=%lu, actual=%lu at %p on %p\n", | |
1439 (unsigned long) size, (unsigned long) object_size, result, | |
1440 (void *) entry); | |
1441 | |
1442 return result; | |
1443 } | |
1444 | |
1445 /* Mark function for strings. */ | |
1446 | |
1447 void | |
1448 gt_ggc_m_S (const void *p) | |
1449 { | |
1450 page_entry *entry; | |
1451 unsigned bit, word; | |
1452 unsigned long mask; | |
1453 unsigned long offset; | |
1454 | |
111 | 1455 if (!p) |
0 | 1456 return; |
1457 | |
111 | 1458 /* Look up the page on which the object is alloced. If it was not |
1459 GC allocated, gracefully bail out. */ | |
1460 entry = safe_lookup_page_table_entry (p); | |
1461 if (!entry) | |
1462 return; | |
0 | 1463 |
1464 /* Calculate the index of the object on the page; this is its bit | |
1465 position in the in_use_p bitmap. Note that because a char* might | |
1466 point to the middle of an object, we need special code here to | |
1467 make sure P points to the start of an object. */ | |
1468 offset = ((const char *) p - entry->page) % object_size_table[entry->order]; | |
1469 if (offset) | |
1470 { | |
1471 /* Here we've seen a char* which does not point to the beginning | |
1472 of an allocated object. We assume it points to the middle of | |
1473 a STRING_CST. */ | |
1474 gcc_assert (offset == offsetof (struct tree_string, str)); | |
1475 p = ((const char *) p) - offset; | |
1476 gt_ggc_mx_lang_tree_node (CONST_CAST (void *, p)); | |
1477 return; | |
1478 } | |
1479 | |
1480 bit = OFFSET_TO_BIT (((const char *) p) - entry->page, entry->order); | |
1481 word = bit / HOST_BITS_PER_LONG; | |
1482 mask = (unsigned long) 1 << (bit % HOST_BITS_PER_LONG); | |
1483 | |
1484 /* If the bit was previously set, skip it. */ | |
1485 if (entry->in_use_p[word] & mask) | |
1486 return; | |
1487 | |
1488 /* Otherwise set it, and decrement the free object count. */ | |
1489 entry->in_use_p[word] |= mask; | |
1490 entry->num_free_objects -= 1; | |
1491 | |
1492 if (GGC_DEBUG_LEVEL >= 4) | |
1493 fprintf (G.debug_file, "Marking %p\n", p); | |
1494 | |
1495 return; | |
1496 } | |
1497 | |
111 | 1498 |
1499 /* User-callable entry points for marking string X. */ | |
1500 | |
1501 void | |
1502 gt_ggc_mx (const char *& x) | |
1503 { | |
1504 gt_ggc_m_S (x); | |
1505 } | |
1506 | |
1507 void | |
1508 gt_ggc_mx (unsigned char *& x) | |
1509 { | |
1510 gt_ggc_m_S (x); | |
1511 } | |
1512 | |
1513 void | |
1514 gt_ggc_mx (unsigned char& x ATTRIBUTE_UNUSED) | |
1515 { | |
1516 } | |
1517 | |
0 | 1518 /* If P is not marked, marks it and return false. Otherwise return true. |
1519 P must have been allocated by the GC allocator; it mustn't point to | |
1520 static objects, stack variables, or memory allocated with malloc. */ | |
1521 | |
1522 int | |
1523 ggc_set_mark (const void *p) | |
1524 { | |
1525 page_entry *entry; | |
1526 unsigned bit, word; | |
1527 unsigned long mask; | |
1528 | |
1529 /* Look up the page on which the object is alloced. If the object | |
1530 wasn't allocated by the collector, we'll probably die. */ | |
1531 entry = lookup_page_table_entry (p); | |
1532 gcc_assert (entry); | |
1533 | |
1534 /* Calculate the index of the object on the page; this is its bit | |
1535 position in the in_use_p bitmap. */ | |
1536 bit = OFFSET_TO_BIT (((const char *) p) - entry->page, entry->order); | |
1537 word = bit / HOST_BITS_PER_LONG; | |
1538 mask = (unsigned long) 1 << (bit % HOST_BITS_PER_LONG); | |
1539 | |
1540 /* If the bit was previously set, skip it. */ | |
1541 if (entry->in_use_p[word] & mask) | |
1542 return 1; | |
1543 | |
1544 /* Otherwise set it, and decrement the free object count. */ | |
1545 entry->in_use_p[word] |= mask; | |
1546 entry->num_free_objects -= 1; | |
1547 | |
1548 if (GGC_DEBUG_LEVEL >= 4) | |
1549 fprintf (G.debug_file, "Marking %p\n", p); | |
1550 | |
1551 return 0; | |
1552 } | |
1553 | |
1554 /* Return 1 if P has been marked, zero otherwise. | |
1555 P must have been allocated by the GC allocator; it mustn't point to | |
1556 static objects, stack variables, or memory allocated with malloc. */ | |
1557 | |
1558 int | |
1559 ggc_marked_p (const void *p) | |
1560 { | |
1561 page_entry *entry; | |
1562 unsigned bit, word; | |
1563 unsigned long mask; | |
1564 | |
1565 /* Look up the page on which the object is alloced. If the object | |
1566 wasn't allocated by the collector, we'll probably die. */ | |
1567 entry = lookup_page_table_entry (p); | |
1568 gcc_assert (entry); | |
1569 | |
1570 /* Calculate the index of the object on the page; this is its bit | |
1571 position in the in_use_p bitmap. */ | |
1572 bit = OFFSET_TO_BIT (((const char *) p) - entry->page, entry->order); | |
1573 word = bit / HOST_BITS_PER_LONG; | |
1574 mask = (unsigned long) 1 << (bit % HOST_BITS_PER_LONG); | |
1575 | |
1576 return (entry->in_use_p[word] & mask) != 0; | |
1577 } | |
1578 | |
1579 /* Return the size of the gc-able object P. */ | |
1580 | |
1581 size_t | |
1582 ggc_get_size (const void *p) | |
1583 { | |
1584 page_entry *pe = lookup_page_table_entry (p); | |
1585 return OBJECT_SIZE (pe->order); | |
1586 } | |
1587 | |
1588 /* Release the memory for object P. */ | |
1589 | |
1590 void | |
1591 ggc_free (void *p) | |
1592 { | |
111 | 1593 if (in_gc) |
1594 return; | |
1595 | |
0 | 1596 page_entry *pe = lookup_page_table_entry (p); |
1597 size_t order = pe->order; | |
1598 size_t size = OBJECT_SIZE (order); | |
1599 | |
111 | 1600 if (GATHER_STATISTICS) |
1601 ggc_free_overhead (p); | |
0 | 1602 |
1603 if (GGC_DEBUG_LEVEL >= 3) | |
1604 fprintf (G.debug_file, | |
1605 "Freeing object, actual size=%lu, at %p on %p\n", | |
1606 (unsigned long) size, p, (void *) pe); | |
1607 | |
1608 #ifdef ENABLE_GC_CHECKING | |
1609 /* Poison the data, to indicate the data is garbage. */ | |
1610 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (p, size)); | |
1611 memset (p, 0xa5, size); | |
1612 #endif | |
1613 /* Let valgrind know the object is free. */ | |
1614 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (p, size)); | |
1615 | |
1616 #ifdef ENABLE_GC_ALWAYS_COLLECT | |
1617 /* In the completely-anal-checking mode, we do *not* immediately free | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1618 the data, but instead verify that the data is *actually* not |
0 | 1619 reachable the next time we collect. */ |
1620 { | |
1621 struct free_object *fo = XNEW (struct free_object); | |
1622 fo->object = p; | |
1623 fo->next = G.free_object_list; | |
1624 G.free_object_list = fo; | |
1625 } | |
1626 #else | |
1627 { | |
1628 unsigned int bit_offset, word, bit; | |
1629 | |
1630 G.allocated -= size; | |
1631 | |
1632 /* Mark the object not-in-use. */ | |
1633 bit_offset = OFFSET_TO_BIT (((const char *) p) - pe->page, order); | |
1634 word = bit_offset / HOST_BITS_PER_LONG; | |
1635 bit = bit_offset % HOST_BITS_PER_LONG; | |
1636 pe->in_use_p[word] &= ~(1UL << bit); | |
1637 | |
1638 if (pe->num_free_objects++ == 0) | |
1639 { | |
1640 page_entry *p, *q; | |
1641 | |
1642 /* If the page is completely full, then it's supposed to | |
1643 be after all pages that aren't. Since we've freed one | |
1644 object from a page that was full, we need to move the | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1645 page to the head of the list. |
0 | 1646 |
1647 PE is the node we want to move. Q is the previous node | |
1648 and P is the next node in the list. */ | |
1649 q = pe->prev; | |
1650 if (q && q->num_free_objects == 0) | |
1651 { | |
1652 p = pe->next; | |
1653 | |
1654 q->next = p; | |
1655 | |
1656 /* If PE was at the end of the list, then Q becomes the | |
1657 new end of the list. If PE was not the end of the | |
1658 list, then we need to update the PREV field for P. */ | |
1659 if (!p) | |
1660 G.page_tails[order] = q; | |
1661 else | |
1662 p->prev = q; | |
1663 | |
1664 /* Move PE to the head of the list. */ | |
1665 pe->next = G.pages[order]; | |
1666 pe->prev = NULL; | |
1667 G.pages[order]->prev = pe; | |
1668 G.pages[order] = pe; | |
1669 } | |
1670 | |
1671 /* Reset the hint bit to point to the only free object. */ | |
1672 pe->next_bit_hint = bit_offset; | |
1673 } | |
1674 } | |
1675 #endif | |
1676 } | |
1677 | |
1678 /* Subroutine of init_ggc which computes the pair of numbers used to | |
1679 perform division by OBJECT_SIZE (order) and fills in inverse_table[]. | |
1680 | |
1681 This algorithm is taken from Granlund and Montgomery's paper | |
1682 "Division by Invariant Integers using Multiplication" | |
1683 (Proc. SIGPLAN PLDI, 1994), section 9 (Exact division by | |
1684 constants). */ | |
1685 | |
1686 static void | |
1687 compute_inverse (unsigned order) | |
1688 { | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1689 size_t size, inv; |
0 | 1690 unsigned int e; |
1691 | |
1692 size = OBJECT_SIZE (order); | |
1693 e = 0; | |
1694 while (size % 2 == 0) | |
1695 { | |
1696 e++; | |
1697 size >>= 1; | |
1698 } | |
1699 | |
1700 inv = size; | |
1701 while (inv * size != 1) | |
1702 inv = inv * (2 - inv*size); | |
1703 | |
1704 DIV_MULT (order) = inv; | |
1705 DIV_SHIFT (order) = e; | |
1706 } | |
1707 | |
1708 /* Initialize the ggc-mmap allocator. */ | |
1709 void | |
1710 init_ggc (void) | |
1711 { | |
111 | 1712 static bool init_p = false; |
0 | 1713 unsigned order; |
1714 | |
111 | 1715 if (init_p) |
1716 return; | |
1717 init_p = true; | |
1718 | |
1719 G.pagesize = getpagesize (); | |
0 | 1720 G.lg_pagesize = exact_log2 (G.pagesize); |
1721 | |
1722 #ifdef HAVE_MMAP_DEV_ZERO | |
1723 G.dev_zero_fd = open ("/dev/zero", O_RDONLY); | |
1724 if (G.dev_zero_fd == -1) | |
1725 internal_error ("open /dev/zero: %m"); | |
1726 #endif | |
1727 | |
1728 #if 0 | |
1729 G.debug_file = fopen ("ggc-mmap.debug", "w"); | |
1730 #else | |
1731 G.debug_file = stdout; | |
1732 #endif | |
1733 | |
1734 #ifdef USING_MMAP | |
1735 /* StunOS has an amazing off-by-one error for the first mmap allocation | |
1736 after fiddling with RLIMIT_STACK. The result, as hard as it is to | |
1737 believe, is an unaligned page allocation, which would cause us to | |
1738 hork badly if we tried to use it. */ | |
1739 { | |
111 | 1740 char *p = alloc_anon (NULL, G.pagesize, true); |
0 | 1741 struct page_entry *e; |
111 | 1742 if ((uintptr_t)p & (G.pagesize - 1)) |
0 | 1743 { |
1744 /* How losing. Discard this one and try another. If we still | |
1745 can't get something useful, give up. */ | |
1746 | |
111 | 1747 p = alloc_anon (NULL, G.pagesize, true); |
1748 gcc_assert (!((uintptr_t)p & (G.pagesize - 1))); | |
0 | 1749 } |
1750 | |
1751 /* We have a good page, might as well hold onto it... */ | |
1752 e = XCNEW (struct page_entry); | |
1753 e->bytes = G.pagesize; | |
1754 e->page = p; | |
1755 e->next = G.free_pages; | |
1756 G.free_pages = e; | |
1757 } | |
1758 #endif | |
1759 | |
1760 /* Initialize the object size table. */ | |
1761 for (order = 0; order < HOST_BITS_PER_PTR; ++order) | |
1762 object_size_table[order] = (size_t) 1 << order; | |
1763 for (order = HOST_BITS_PER_PTR; order < NUM_ORDERS; ++order) | |
1764 { | |
1765 size_t s = extra_order_size_table[order - HOST_BITS_PER_PTR]; | |
1766 | |
1767 /* If S is not a multiple of the MAX_ALIGNMENT, then round it up | |
1768 so that we're sure of getting aligned memory. */ | |
1769 s = ROUND_UP (s, MAX_ALIGNMENT); | |
1770 object_size_table[order] = s; | |
1771 } | |
1772 | |
1773 /* Initialize the objects-per-page and inverse tables. */ | |
1774 for (order = 0; order < NUM_ORDERS; ++order) | |
1775 { | |
1776 objects_per_page_table[order] = G.pagesize / OBJECT_SIZE (order); | |
1777 if (objects_per_page_table[order] == 0) | |
1778 objects_per_page_table[order] = 1; | |
1779 compute_inverse (order); | |
1780 } | |
1781 | |
1782 /* Reset the size_lookup array to put appropriately sized objects in | |
1783 the special orders. All objects bigger than the previous power | |
1784 of two, but no greater than the special size, should go in the | |
1785 new order. */ | |
1786 for (order = HOST_BITS_PER_PTR; order < NUM_ORDERS; ++order) | |
1787 { | |
1788 int o; | |
1789 int i; | |
1790 | |
1791 i = OBJECT_SIZE (order); | |
1792 if (i >= NUM_SIZE_LOOKUP) | |
1793 continue; | |
1794 | |
1795 for (o = size_lookup[i]; o == size_lookup [i]; --i) | |
1796 size_lookup[i] = order; | |
1797 } | |
1798 | |
1799 G.depth_in_use = 0; | |
1800 G.depth_max = 10; | |
1801 G.depth = XNEWVEC (unsigned int, G.depth_max); | |
1802 | |
1803 G.by_depth_in_use = 0; | |
1804 G.by_depth_max = INITIAL_PTE_COUNT; | |
1805 G.by_depth = XNEWVEC (page_entry *, G.by_depth_max); | |
1806 G.save_in_use = XNEWVEC (unsigned long *, G.by_depth_max); | |
111 | 1807 |
1808 /* Allocate space for the depth 0 finalizers. */ | |
1809 G.finalizers.safe_push (vNULL); | |
1810 G.vec_finalizers.safe_push (vNULL); | |
1811 gcc_assert (G.finalizers.length() == 1); | |
0 | 1812 } |
1813 | |
1814 /* Merge the SAVE_IN_USE_P and IN_USE_P arrays in P so that IN_USE_P | |
1815 reflects reality. Recalculate NUM_FREE_OBJECTS as well. */ | |
1816 | |
1817 static void | |
1818 ggc_recalculate_in_use_p (page_entry *p) | |
1819 { | |
1820 unsigned int i; | |
1821 size_t num_objects; | |
1822 | |
1823 /* Because the past-the-end bit in in_use_p is always set, we | |
1824 pretend there is one additional object. */ | |
1825 num_objects = OBJECTS_IN_PAGE (p) + 1; | |
1826 | |
1827 /* Reset the free object count. */ | |
1828 p->num_free_objects = num_objects; | |
1829 | |
1830 /* Combine the IN_USE_P and SAVE_IN_USE_P arrays. */ | |
1831 for (i = 0; | |
1832 i < CEIL (BITMAP_SIZE (num_objects), | |
1833 sizeof (*p->in_use_p)); | |
1834 ++i) | |
1835 { | |
1836 unsigned long j; | |
1837 | |
1838 /* Something is in use if it is marked, or if it was in use in a | |
1839 context further down the context stack. */ | |
1840 p->in_use_p[i] |= save_in_use_p (p)[i]; | |
1841 | |
1842 /* Decrement the free object count for every object allocated. */ | |
1843 for (j = p->in_use_p[i]; j; j >>= 1) | |
1844 p->num_free_objects -= (j & 1); | |
1845 } | |
1846 | |
1847 gcc_assert (p->num_free_objects < num_objects); | |
1848 } | |
1849 | |
1850 /* Unmark all objects. */ | |
1851 | |
1852 static void | |
1853 clear_marks (void) | |
1854 { | |
1855 unsigned order; | |
1856 | |
1857 for (order = 2; order < NUM_ORDERS; order++) | |
1858 { | |
1859 page_entry *p; | |
1860 | |
1861 for (p = G.pages[order]; p != NULL; p = p->next) | |
1862 { | |
1863 size_t num_objects = OBJECTS_IN_PAGE (p); | |
1864 size_t bitmap_size = BITMAP_SIZE (num_objects + 1); | |
1865 | |
1866 /* The data should be page-aligned. */ | |
111 | 1867 gcc_assert (!((uintptr_t) p->page & (G.pagesize - 1))); |
0 | 1868 |
1869 /* Pages that aren't in the topmost context are not collected; | |
1870 nevertheless, we need their in-use bit vectors to store GC | |
1871 marks. So, back them up first. */ | |
1872 if (p->context_depth < G.context_depth) | |
1873 { | |
1874 if (! save_in_use_p (p)) | |
1875 save_in_use_p (p) = XNEWVAR (unsigned long, bitmap_size); | |
1876 memcpy (save_in_use_p (p), p->in_use_p, bitmap_size); | |
1877 } | |
1878 | |
1879 /* Reset reset the number of free objects and clear the | |
1880 in-use bits. These will be adjusted by mark_obj. */ | |
1881 p->num_free_objects = num_objects; | |
1882 memset (p->in_use_p, 0, bitmap_size); | |
1883 | |
1884 /* Make sure the one-past-the-end bit is always set. */ | |
1885 p->in_use_p[num_objects / HOST_BITS_PER_LONG] | |
1886 = ((unsigned long) 1 << (num_objects % HOST_BITS_PER_LONG)); | |
1887 } | |
1888 } | |
1889 } | |
1890 | |
111 | 1891 /* Check if any blocks with a registered finalizer have become unmarked. If so |
1892 run the finalizer and unregister it because the block is about to be freed. | |
1893 Note that no garantee is made about what order finalizers will run in so | |
1894 touching other objects in gc memory is extremely unwise. */ | |
1895 | |
1896 static void | |
1897 ggc_handle_finalizers () | |
1898 { | |
1899 unsigned dlen = G.finalizers.length(); | |
1900 for (unsigned d = G.context_depth; d < dlen; ++d) | |
1901 { | |
1902 vec<finalizer> &v = G.finalizers[d]; | |
1903 unsigned length = v.length (); | |
1904 for (unsigned int i = 0; i < length;) | |
1905 { | |
1906 finalizer &f = v[i]; | |
1907 if (!ggc_marked_p (f.addr ())) | |
1908 { | |
1909 f.call (); | |
1910 v.unordered_remove (i); | |
1911 length--; | |
1912 } | |
1913 else | |
1914 i++; | |
1915 } | |
1916 } | |
1917 | |
1918 gcc_assert (dlen == G.vec_finalizers.length()); | |
1919 for (unsigned d = G.context_depth; d < dlen; ++d) | |
1920 { | |
1921 vec<vec_finalizer> &vv = G.vec_finalizers[d]; | |
1922 unsigned length = vv.length (); | |
1923 for (unsigned int i = 0; i < length;) | |
1924 { | |
1925 vec_finalizer &f = vv[i]; | |
1926 if (!ggc_marked_p (f.addr ())) | |
1927 { | |
1928 f.call (); | |
1929 vv.unordered_remove (i); | |
1930 length--; | |
1931 } | |
1932 else | |
1933 i++; | |
1934 } | |
1935 } | |
1936 } | |
1937 | |
0 | 1938 /* Free all empty pages. Partially empty pages need no attention |
1939 because the `mark' bit doubles as an `unused' bit. */ | |
1940 | |
1941 static void | |
1942 sweep_pages (void) | |
1943 { | |
1944 unsigned order; | |
1945 | |
1946 for (order = 2; order < NUM_ORDERS; order++) | |
1947 { | |
1948 /* The last page-entry to consider, regardless of entries | |
1949 placed at the end of the list. */ | |
1950 page_entry * const last = G.page_tails[order]; | |
1951 | |
1952 size_t num_objects; | |
1953 size_t live_objects; | |
1954 page_entry *p, *previous; | |
1955 int done; | |
1956 | |
1957 p = G.pages[order]; | |
1958 if (p == NULL) | |
1959 continue; | |
1960 | |
1961 previous = NULL; | |
1962 do | |
1963 { | |
1964 page_entry *next = p->next; | |
1965 | |
1966 /* Loop until all entries have been examined. */ | |
1967 done = (p == last); | |
1968 | |
1969 num_objects = OBJECTS_IN_PAGE (p); | |
1970 | |
1971 /* Add all live objects on this page to the count of | |
1972 allocated memory. */ | |
1973 live_objects = num_objects - p->num_free_objects; | |
1974 | |
1975 G.allocated += OBJECT_SIZE (order) * live_objects; | |
1976 | |
1977 /* Only objects on pages in the topmost context should get | |
1978 collected. */ | |
1979 if (p->context_depth < G.context_depth) | |
1980 ; | |
1981 | |
1982 /* Remove the page if it's empty. */ | |
1983 else if (live_objects == 0) | |
1984 { | |
1985 /* If P was the first page in the list, then NEXT | |
1986 becomes the new first page in the list, otherwise | |
1987 splice P out of the forward pointers. */ | |
1988 if (! previous) | |
1989 G.pages[order] = next; | |
1990 else | |
1991 previous->next = next; | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1992 |
0 | 1993 /* Splice P out of the back pointers too. */ |
1994 if (next) | |
1995 next->prev = previous; | |
1996 | |
1997 /* Are we removing the last element? */ | |
1998 if (p == G.page_tails[order]) | |
1999 G.page_tails[order] = previous; | |
2000 free_page (p); | |
2001 p = previous; | |
2002 } | |
2003 | |
2004 /* If the page is full, move it to the end. */ | |
2005 else if (p->num_free_objects == 0) | |
2006 { | |
2007 /* Don't move it if it's already at the end. */ | |
2008 if (p != G.page_tails[order]) | |
2009 { | |
2010 /* Move p to the end of the list. */ | |
2011 p->next = NULL; | |
2012 p->prev = G.page_tails[order]; | |
2013 G.page_tails[order]->next = p; | |
2014 | |
2015 /* Update the tail pointer... */ | |
2016 G.page_tails[order] = p; | |
2017 | |
2018 /* ... and the head pointer, if necessary. */ | |
2019 if (! previous) | |
2020 G.pages[order] = next; | |
2021 else | |
2022 previous->next = next; | |
2023 | |
2024 /* And update the backpointer in NEXT if necessary. */ | |
2025 if (next) | |
2026 next->prev = previous; | |
2027 | |
2028 p = previous; | |
2029 } | |
2030 } | |
2031 | |
2032 /* If we've fallen through to here, it's a page in the | |
2033 topmost context that is neither full nor empty. Such a | |
2034 page must precede pages at lesser context depth in the | |
2035 list, so move it to the head. */ | |
2036 else if (p != G.pages[order]) | |
2037 { | |
2038 previous->next = p->next; | |
2039 | |
2040 /* Update the backchain in the next node if it exists. */ | |
2041 if (p->next) | |
2042 p->next->prev = previous; | |
2043 | |
2044 /* Move P to the head of the list. */ | |
2045 p->next = G.pages[order]; | |
2046 p->prev = NULL; | |
2047 G.pages[order]->prev = p; | |
2048 | |
2049 /* Update the head pointer. */ | |
2050 G.pages[order] = p; | |
2051 | |
2052 /* Are we moving the last element? */ | |
2053 if (G.page_tails[order] == p) | |
2054 G.page_tails[order] = previous; | |
2055 p = previous; | |
2056 } | |
2057 | |
2058 previous = p; | |
2059 p = next; | |
2060 } | |
2061 while (! done); | |
2062 | |
2063 /* Now, restore the in_use_p vectors for any pages from contexts | |
2064 other than the current one. */ | |
2065 for (p = G.pages[order]; p; p = p->next) | |
2066 if (p->context_depth != G.context_depth) | |
2067 ggc_recalculate_in_use_p (p); | |
2068 } | |
2069 } | |
2070 | |
2071 #ifdef ENABLE_GC_CHECKING | |
2072 /* Clobber all free objects. */ | |
2073 | |
2074 static void | |
2075 poison_pages (void) | |
2076 { | |
2077 unsigned order; | |
2078 | |
2079 for (order = 2; order < NUM_ORDERS; order++) | |
2080 { | |
2081 size_t size = OBJECT_SIZE (order); | |
2082 page_entry *p; | |
2083 | |
2084 for (p = G.pages[order]; p != NULL; p = p->next) | |
2085 { | |
2086 size_t num_objects; | |
2087 size_t i; | |
2088 | |
2089 if (p->context_depth != G.context_depth) | |
2090 /* Since we don't do any collection for pages in pushed | |
2091 contexts, there's no need to do any poisoning. And | |
2092 besides, the IN_USE_P array isn't valid until we pop | |
2093 contexts. */ | |
2094 continue; | |
2095 | |
2096 num_objects = OBJECTS_IN_PAGE (p); | |
2097 for (i = 0; i < num_objects; i++) | |
2098 { | |
2099 size_t word, bit; | |
2100 word = i / HOST_BITS_PER_LONG; | |
2101 bit = i % HOST_BITS_PER_LONG; | |
2102 if (((p->in_use_p[word] >> bit) & 1) == 0) | |
2103 { | |
2104 char *object = p->page + i * size; | |
2105 | |
2106 /* Keep poison-by-write when we expect to use Valgrind, | |
2107 so the exact same memory semantics is kept, in case | |
2108 there are memory errors. We override this request | |
2109 below. */ | |
2110 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (object, | |
2111 size)); | |
2112 memset (object, 0xa5, size); | |
2113 | |
2114 /* Drop the handle to avoid handle leak. */ | |
2115 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (object, size)); | |
2116 } | |
2117 } | |
2118 } | |
2119 } | |
2120 } | |
2121 #else | |
2122 #define poison_pages() | |
2123 #endif | |
2124 | |
2125 #ifdef ENABLE_GC_ALWAYS_COLLECT | |
2126 /* Validate that the reportedly free objects actually are. */ | |
2127 | |
2128 static void | |
2129 validate_free_objects (void) | |
2130 { | |
2131 struct free_object *f, *next, *still_free = NULL; | |
2132 | |
2133 for (f = G.free_object_list; f ; f = next) | |
2134 { | |
2135 page_entry *pe = lookup_page_table_entry (f->object); | |
2136 size_t bit, word; | |
2137 | |
2138 bit = OFFSET_TO_BIT ((char *)f->object - pe->page, pe->order); | |
2139 word = bit / HOST_BITS_PER_LONG; | |
2140 bit = bit % HOST_BITS_PER_LONG; | |
2141 next = f->next; | |
2142 | |
2143 /* Make certain it isn't visible from any root. Notice that we | |
2144 do this check before sweep_pages merges save_in_use_p. */ | |
2145 gcc_assert (!(pe->in_use_p[word] & (1UL << bit))); | |
2146 | |
2147 /* If the object comes from an outer context, then retain the | |
2148 free_object entry, so that we can verify that the address | |
2149 isn't live on the stack in some outer context. */ | |
2150 if (pe->context_depth != G.context_depth) | |
2151 { | |
2152 f->next = still_free; | |
2153 still_free = f; | |
2154 } | |
2155 else | |
2156 free (f); | |
2157 } | |
2158 | |
2159 G.free_object_list = still_free; | |
2160 } | |
2161 #else | |
2162 #define validate_free_objects() | |
2163 #endif | |
2164 | |
2165 /* Top level mark-and-sweep routine. */ | |
2166 | |
2167 void | |
2168 ggc_collect (void) | |
2169 { | |
2170 /* Avoid frequent unnecessary work by skipping collection if the | |
2171 total allocations haven't expanded much since the last | |
2172 collection. */ | |
2173 float allocated_last_gc = | |
2174 MAX (G.allocated_last_gc, (size_t)PARAM_VALUE (GGC_MIN_HEAPSIZE) * 1024); | |
2175 | |
2176 float min_expand = allocated_last_gc * PARAM_VALUE (GGC_MIN_EXPAND) / 100; | |
2177 if (G.allocated < allocated_last_gc + min_expand && !ggc_force_collect) | |
2178 return; | |
2179 | |
2180 timevar_push (TV_GC); | |
2181 if (!quiet_flag) | |
2182 fprintf (stderr, " {GC %luk -> ", (unsigned long) G.allocated / 1024); | |
2183 if (GGC_DEBUG_LEVEL >= 2) | |
2184 fprintf (G.debug_file, "BEGIN COLLECTING\n"); | |
2185 | |
2186 /* Zero the total allocated bytes. This will be recalculated in the | |
2187 sweep phase. */ | |
2188 G.allocated = 0; | |
2189 | |
2190 /* Release the pages we freed the last time we collected, but didn't | |
2191 reuse in the interim. */ | |
2192 release_pages (); | |
2193 | |
2194 /* Indicate that we've seen collections at this context depth. */ | |
2195 G.context_depth_collections = ((unsigned long)1 << (G.context_depth + 1)) - 1; | |
2196 | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2197 invoke_plugin_callbacks (PLUGIN_GGC_START, NULL); |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2198 |
111 | 2199 in_gc = true; |
0 | 2200 clear_marks (); |
2201 ggc_mark_roots (); | |
111 | 2202 ggc_handle_finalizers (); |
2203 | |
2204 if (GATHER_STATISTICS) | |
2205 ggc_prune_overhead_list (); | |
2206 | |
0 | 2207 poison_pages (); |
2208 validate_free_objects (); | |
2209 sweep_pages (); | |
2210 | |
111 | 2211 in_gc = false; |
0 | 2212 G.allocated_last_gc = G.allocated; |
2213 | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2214 invoke_plugin_callbacks (PLUGIN_GGC_END, NULL); |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2215 |
0 | 2216 timevar_pop (TV_GC); |
2217 | |
2218 if (!quiet_flag) | |
2219 fprintf (stderr, "%luk}", (unsigned long) G.allocated / 1024); | |
2220 if (GGC_DEBUG_LEVEL >= 2) | |
2221 fprintf (G.debug_file, "END COLLECTING\n"); | |
2222 } | |
2223 | |
111 | 2224 /* Assume that all GGC memory is reachable and grow the limits for next collection. |
2225 With checking, trigger GGC so -Q compilation outputs how much of memory really is | |
2226 reachable. */ | |
2227 | |
2228 void | |
2229 ggc_grow (void) | |
2230 { | |
2231 if (!flag_checking) | |
2232 G.allocated_last_gc = MAX (G.allocated_last_gc, | |
2233 G.allocated); | |
2234 else | |
2235 ggc_collect (); | |
2236 if (!quiet_flag) | |
2237 fprintf (stderr, " {GC start %luk} ", (unsigned long) G.allocated / 1024); | |
2238 } | |
2239 | |
0 | 2240 /* Print allocation statistics. */ |
2241 #define SCALE(x) ((unsigned long) ((x) < 1024*10 \ | |
2242 ? (x) \ | |
2243 : ((x) < 1024*1024*10 \ | |
2244 ? (x) / 1024 \ | |
2245 : (x) / (1024*1024)))) | |
2246 #define STAT_LABEL(x) ((x) < 1024*10 ? ' ' : ((x) < 1024*1024*10 ? 'k' : 'M')) | |
2247 | |
2248 void | |
2249 ggc_print_statistics (void) | |
2250 { | |
2251 struct ggc_statistics stats; | |
2252 unsigned int i; | |
2253 size_t total_overhead = 0; | |
2254 | |
2255 /* Clear the statistics. */ | |
2256 memset (&stats, 0, sizeof (stats)); | |
2257 | |
2258 /* Make sure collection will really occur. */ | |
2259 G.allocated_last_gc = 0; | |
2260 | |
2261 /* Collect and print the statistics common across collectors. */ | |
2262 ggc_print_common_statistics (stderr, &stats); | |
2263 | |
2264 /* Release free pages so that we will not count the bytes allocated | |
2265 there as part of the total allocated memory. */ | |
2266 release_pages (); | |
2267 | |
2268 /* Collect some information about the various sizes of | |
2269 allocation. */ | |
2270 fprintf (stderr, | |
2271 "Memory still allocated at the end of the compilation process\n"); | |
111 | 2272 fprintf (stderr, "%-8s %10s %10s %10s\n", |
0 | 2273 "Size", "Allocated", "Used", "Overhead"); |
2274 for (i = 0; i < NUM_ORDERS; ++i) | |
2275 { | |
2276 page_entry *p; | |
2277 size_t allocated; | |
2278 size_t in_use; | |
2279 size_t overhead; | |
2280 | |
2281 /* Skip empty entries. */ | |
2282 if (!G.pages[i]) | |
2283 continue; | |
2284 | |
2285 overhead = allocated = in_use = 0; | |
2286 | |
2287 /* Figure out the total number of bytes allocated for objects of | |
2288 this size, and how many of them are actually in use. Also figure | |
2289 out how much memory the page table is using. */ | |
2290 for (p = G.pages[i]; p; p = p->next) | |
2291 { | |
2292 allocated += p->bytes; | |
2293 in_use += | |
2294 (OBJECTS_IN_PAGE (p) - p->num_free_objects) * OBJECT_SIZE (i); | |
2295 | |
2296 overhead += (sizeof (page_entry) - sizeof (long) | |
2297 + BITMAP_SIZE (OBJECTS_IN_PAGE (p) + 1)); | |
2298 } | |
111 | 2299 fprintf (stderr, "%-8lu %10lu%c %10lu%c %10lu%c\n", |
0 | 2300 (unsigned long) OBJECT_SIZE (i), |
2301 SCALE (allocated), STAT_LABEL (allocated), | |
2302 SCALE (in_use), STAT_LABEL (in_use), | |
2303 SCALE (overhead), STAT_LABEL (overhead)); | |
2304 total_overhead += overhead; | |
2305 } | |
111 | 2306 fprintf (stderr, "%-8s %10lu%c %10lu%c %10lu%c\n", "Total", |
0 | 2307 SCALE (G.bytes_mapped), STAT_LABEL (G.bytes_mapped), |
111 | 2308 SCALE (G.allocated), STAT_LABEL (G.allocated), |
0 | 2309 SCALE (total_overhead), STAT_LABEL (total_overhead)); |
2310 | |
111 | 2311 if (GATHER_STATISTICS) |
2312 { | |
2313 fprintf (stderr, "\nTotal allocations and overheads during " | |
2314 "the compilation process\n"); | |
2315 | |
2316 fprintf (stderr, "Total Overhead: %10" | |
2317 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_overhead); | |
2318 fprintf (stderr, "Total Allocated: %10" | |
2319 HOST_LONG_LONG_FORMAT "d\n", | |
2320 G.stats.total_allocated); | |
2321 | |
2322 fprintf (stderr, "Total Overhead under 32B: %10" | |
2323 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_overhead_under32); | |
2324 fprintf (stderr, "Total Allocated under 32B: %10" | |
2325 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_allocated_under32); | |
2326 fprintf (stderr, "Total Overhead under 64B: %10" | |
2327 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_overhead_under64); | |
2328 fprintf (stderr, "Total Allocated under 64B: %10" | |
2329 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_allocated_under64); | |
2330 fprintf (stderr, "Total Overhead under 128B: %10" | |
2331 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_overhead_under128); | |
2332 fprintf (stderr, "Total Allocated under 128B: %10" | |
2333 HOST_LONG_LONG_FORMAT "d\n", G.stats.total_allocated_under128); | |
2334 | |
2335 for (i = 0; i < NUM_ORDERS; i++) | |
2336 if (G.stats.total_allocated_per_order[i]) | |
2337 { | |
2338 fprintf (stderr, "Total Overhead page size %9lu: %10" | |
2339 HOST_LONG_LONG_FORMAT "d\n", | |
2340 (unsigned long) OBJECT_SIZE (i), | |
2341 G.stats.total_overhead_per_order[i]); | |
2342 fprintf (stderr, "Total Allocated page size %9lu: %10" | |
2343 HOST_LONG_LONG_FORMAT "d\n", | |
2344 (unsigned long) OBJECT_SIZE (i), | |
2345 G.stats.total_allocated_per_order[i]); | |
2346 } | |
0 | 2347 } |
2348 } | |
2349 | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2350 struct ggc_pch_ondisk |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2351 { |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2352 unsigned totals[NUM_ORDERS]; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2353 }; |
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2354 |
0 | 2355 struct ggc_pch_data |
2356 { | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2357 struct ggc_pch_ondisk d; |
111 | 2358 uintptr_t base[NUM_ORDERS]; |
0 | 2359 size_t written[NUM_ORDERS]; |
2360 }; | |
2361 | |
2362 struct ggc_pch_data * | |
2363 init_ggc_pch (void) | |
2364 { | |
2365 return XCNEW (struct ggc_pch_data); | |
2366 } | |
2367 | |
2368 void | |
2369 ggc_pch_count_object (struct ggc_pch_data *d, void *x ATTRIBUTE_UNUSED, | |
111 | 2370 size_t size, bool is_string ATTRIBUTE_UNUSED) |
0 | 2371 { |
2372 unsigned order; | |
2373 | |
2374 if (size < NUM_SIZE_LOOKUP) | |
2375 order = size_lookup[size]; | |
2376 else | |
2377 { | |
2378 order = 10; | |
2379 while (size > OBJECT_SIZE (order)) | |
2380 order++; | |
2381 } | |
2382 | |
2383 d->d.totals[order]++; | |
2384 } | |
2385 | |
2386 size_t | |
2387 ggc_pch_total_size (struct ggc_pch_data *d) | |
2388 { | |
2389 size_t a = 0; | |
2390 unsigned i; | |
2391 | |
2392 for (i = 0; i < NUM_ORDERS; i++) | |
111 | 2393 a += PAGE_ALIGN (d->d.totals[i] * OBJECT_SIZE (i)); |
0 | 2394 return a; |
2395 } | |
2396 | |
2397 void | |
2398 ggc_pch_this_base (struct ggc_pch_data *d, void *base) | |
2399 { | |
111 | 2400 uintptr_t a = (uintptr_t) base; |
0 | 2401 unsigned i; |
2402 | |
2403 for (i = 0; i < NUM_ORDERS; i++) | |
2404 { | |
2405 d->base[i] = a; | |
111 | 2406 a += PAGE_ALIGN (d->d.totals[i] * OBJECT_SIZE (i)); |
0 | 2407 } |
2408 } | |
2409 | |
2410 | |
2411 char * | |
2412 ggc_pch_alloc_object (struct ggc_pch_data *d, void *x ATTRIBUTE_UNUSED, | |
111 | 2413 size_t size, bool is_string ATTRIBUTE_UNUSED) |
0 | 2414 { |
2415 unsigned order; | |
2416 char *result; | |
2417 | |
2418 if (size < NUM_SIZE_LOOKUP) | |
2419 order = size_lookup[size]; | |
2420 else | |
2421 { | |
2422 order = 10; | |
2423 while (size > OBJECT_SIZE (order)) | |
2424 order++; | |
2425 } | |
2426 | |
2427 result = (char *) d->base[order]; | |
2428 d->base[order] += OBJECT_SIZE (order); | |
2429 return result; | |
2430 } | |
2431 | |
2432 void | |
2433 ggc_pch_prepare_write (struct ggc_pch_data *d ATTRIBUTE_UNUSED, | |
2434 FILE *f ATTRIBUTE_UNUSED) | |
2435 { | |
2436 /* Nothing to do. */ | |
2437 } | |
2438 | |
2439 void | |
111 | 2440 ggc_pch_write_object (struct ggc_pch_data *d, |
0 | 2441 FILE *f, void *x, void *newx ATTRIBUTE_UNUSED, |
2442 size_t size, bool is_string ATTRIBUTE_UNUSED) | |
2443 { | |
2444 unsigned order; | |
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2445 static const char emptyBytes[256] = { 0 }; |
0 | 2446 |
2447 if (size < NUM_SIZE_LOOKUP) | |
2448 order = size_lookup[size]; | |
2449 else | |
2450 { | |
2451 order = 10; | |
2452 while (size > OBJECT_SIZE (order)) | |
2453 order++; | |
2454 } | |
2455 | |
2456 if (fwrite (x, size, 1, f) != 1) | |
111 | 2457 fatal_error (input_location, "can%'t write PCH file: %m"); |
0 | 2458 |
2459 /* If SIZE is not the same as OBJECT_SIZE(order), then we need to pad the | |
2460 object out to OBJECT_SIZE(order). This happens for strings. */ | |
2461 | |
2462 if (size != OBJECT_SIZE (order)) | |
2463 { | |
111 | 2464 unsigned padding = OBJECT_SIZE (order) - size; |
0 | 2465 |
2466 /* To speed small writes, we use a nulled-out array that's larger | |
2467 than most padding requests as the source for our null bytes. This | |
2468 permits us to do the padding with fwrite() rather than fseek(), and | |
2469 limits the chance the OS may try to flush any outstanding writes. */ | |
111 | 2470 if (padding <= sizeof (emptyBytes)) |
0 | 2471 { |
2472 if (fwrite (emptyBytes, 1, padding, f) != padding) | |
111 | 2473 fatal_error (input_location, "can%'t write PCH file"); |
0 | 2474 } |
2475 else | |
2476 { | |
2477 /* Larger than our buffer? Just default to fseek. */ | |
2478 if (fseek (f, padding, SEEK_CUR) != 0) | |
111 | 2479 fatal_error (input_location, "can%'t write PCH file"); |
0 | 2480 } |
2481 } | |
2482 | |
2483 d->written[order]++; | |
2484 if (d->written[order] == d->d.totals[order] | |
2485 && fseek (f, ROUND_UP_VALUE (d->d.totals[order] * OBJECT_SIZE (order), | |
2486 G.pagesize), | |
2487 SEEK_CUR) != 0) | |
111 | 2488 fatal_error (input_location, "can%'t write PCH file: %m"); |
0 | 2489 } |
2490 | |
2491 void | |
2492 ggc_pch_finish (struct ggc_pch_data *d, FILE *f) | |
2493 { | |
2494 if (fwrite (&d->d, sizeof (d->d), 1, f) != 1) | |
111 | 2495 fatal_error (input_location, "can%'t write PCH file: %m"); |
0 | 2496 free (d); |
2497 } | |
2498 | |
2499 /* Move the PCH PTE entries just added to the end of by_depth, to the | |
2500 front. */ | |
2501 | |
2502 static void | |
2503 move_ptes_to_front (int count_old_page_tables, int count_new_page_tables) | |
2504 { | |
2505 /* First, we swap the new entries to the front of the varrays. */ | |
2506 page_entry **new_by_depth; | |
2507 unsigned long **new_save_in_use; | |
2508 | |
2509 new_by_depth = XNEWVEC (page_entry *, G.by_depth_max); | |
2510 new_save_in_use = XNEWVEC (unsigned long *, G.by_depth_max); | |
2511 | |
2512 memcpy (&new_by_depth[0], | |
2513 &G.by_depth[count_old_page_tables], | |
2514 count_new_page_tables * sizeof (void *)); | |
2515 memcpy (&new_by_depth[count_new_page_tables], | |
2516 &G.by_depth[0], | |
2517 count_old_page_tables * sizeof (void *)); | |
2518 memcpy (&new_save_in_use[0], | |
2519 &G.save_in_use[count_old_page_tables], | |
2520 count_new_page_tables * sizeof (void *)); | |
2521 memcpy (&new_save_in_use[count_new_page_tables], | |
2522 &G.save_in_use[0], | |
2523 count_old_page_tables * sizeof (void *)); | |
2524 | |
2525 free (G.by_depth); | |
2526 free (G.save_in_use); | |
2527 | |
2528 G.by_depth = new_by_depth; | |
2529 G.save_in_use = new_save_in_use; | |
2530 | |
2531 /* Now update all the index_by_depth fields. */ | |
111 | 2532 for (unsigned i = G.by_depth_in_use; i--;) |
0 | 2533 { |
111 | 2534 page_entry *p = G.by_depth[i]; |
2535 p->index_by_depth = i; | |
0 | 2536 } |
2537 | |
2538 /* And last, we update the depth pointers in G.depth. The first | |
2539 entry is already 0, and context 0 entries always start at index | |
2540 0, so there is nothing to update in the first slot. We need a | |
2541 second slot, only if we have old ptes, and if we do, they start | |
2542 at index count_new_page_tables. */ | |
2543 if (count_old_page_tables) | |
2544 push_depth (count_new_page_tables); | |
2545 } | |
2546 | |
2547 void | |
2548 ggc_pch_read (FILE *f, void *addr) | |
2549 { | |
2550 struct ggc_pch_ondisk d; | |
2551 unsigned i; | |
2552 char *offs = (char *) addr; | |
2553 unsigned long count_old_page_tables; | |
2554 unsigned long count_new_page_tables; | |
2555 | |
2556 count_old_page_tables = G.by_depth_in_use; | |
2557 | |
2558 /* We've just read in a PCH file. So, every object that used to be | |
2559 allocated is now free. */ | |
2560 clear_marks (); | |
2561 #ifdef ENABLE_GC_CHECKING | |
2562 poison_pages (); | |
2563 #endif | |
2564 /* Since we free all the allocated objects, the free list becomes | |
2565 useless. Validate it now, which will also clear it. */ | |
111 | 2566 validate_free_objects (); |
0 | 2567 |
2568 /* No object read from a PCH file should ever be freed. So, set the | |
2569 context depth to 1, and set the depth of all the currently-allocated | |
2570 pages to be 1 too. PCH pages will have depth 0. */ | |
2571 gcc_assert (!G.context_depth); | |
2572 G.context_depth = 1; | |
111 | 2573 /* Allocate space for the depth 1 finalizers. */ |
2574 G.finalizers.safe_push (vNULL); | |
2575 G.vec_finalizers.safe_push (vNULL); | |
2576 gcc_assert (G.finalizers.length() == 2); | |
0 | 2577 for (i = 0; i < NUM_ORDERS; i++) |
2578 { | |
2579 page_entry *p; | |
2580 for (p = G.pages[i]; p != NULL; p = p->next) | |
2581 p->context_depth = G.context_depth; | |
2582 } | |
2583 | |
2584 /* Allocate the appropriate page-table entries for the pages read from | |
2585 the PCH file. */ | |
2586 if (fread (&d, sizeof (d), 1, f) != 1) | |
111 | 2587 fatal_error (input_location, "can%'t read PCH file: %m"); |
0 | 2588 |
2589 for (i = 0; i < NUM_ORDERS; i++) | |
2590 { | |
2591 struct page_entry *entry; | |
2592 char *pte; | |
2593 size_t bytes; | |
2594 size_t num_objs; | |
2595 size_t j; | |
2596 | |
2597 if (d.totals[i] == 0) | |
2598 continue; | |
2599 | |
111 | 2600 bytes = PAGE_ALIGN (d.totals[i] * OBJECT_SIZE (i)); |
0 | 2601 num_objs = bytes / OBJECT_SIZE (i); |
2602 entry = XCNEWVAR (struct page_entry, (sizeof (struct page_entry) | |
2603 - sizeof (long) | |
2604 + BITMAP_SIZE (num_objs + 1))); | |
2605 entry->bytes = bytes; | |
2606 entry->page = offs; | |
2607 entry->context_depth = 0; | |
2608 offs += bytes; | |
2609 entry->num_free_objects = 0; | |
2610 entry->order = i; | |
2611 | |
2612 for (j = 0; | |
2613 j + HOST_BITS_PER_LONG <= num_objs + 1; | |
2614 j += HOST_BITS_PER_LONG) | |
2615 entry->in_use_p[j / HOST_BITS_PER_LONG] = -1; | |
2616 for (; j < num_objs + 1; j++) | |
2617 entry->in_use_p[j / HOST_BITS_PER_LONG] | |
2618 |= 1L << (j % HOST_BITS_PER_LONG); | |
2619 | |
2620 for (pte = entry->page; | |
2621 pte < entry->page + entry->bytes; | |
2622 pte += G.pagesize) | |
2623 set_page_table_entry (pte, entry); | |
2624 | |
2625 if (G.page_tails[i] != NULL) | |
2626 G.page_tails[i]->next = entry; | |
2627 else | |
2628 G.pages[i] = entry; | |
2629 G.page_tails[i] = entry; | |
2630 | |
2631 /* We start off by just adding all the new information to the | |
2632 end of the varrays, later, we will move the new information | |
2633 to the front of the varrays, as the PCH page tables are at | |
2634 context 0. */ | |
2635 push_by_depth (entry, 0); | |
2636 } | |
2637 | |
2638 /* Now, we update the various data structures that speed page table | |
2639 handling. */ | |
2640 count_new_page_tables = G.by_depth_in_use - count_old_page_tables; | |
2641 | |
2642 move_ptes_to_front (count_old_page_tables, count_new_page_tables); | |
2643 | |
2644 /* Update the statistics. */ | |
2645 G.allocated = G.allocated_last_gc = offs - (char *)addr; | |
2646 } |