Mercurial > hg > CbC > CbC_gcc
annotate gcc/ggc-page.c @ 58:3aaf117db171
error at dwarf2out.c
| author | ryoma <e075725@ie.u-ryukyu.ac.jp> |
|---|---|
| date | Mon, 15 Feb 2010 14:58:24 +0900 |
| parents | 77e2b8dfacca |
| children | f6334be47118 |
| rev | line source |
|---|---|
| 0 | 1 /* "Bag-of-pages" garbage collector for the GNU compiler. |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2 Copyright (C) 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2007, 2008, 2009 |
| 0 | 3 Free Software Foundation, Inc. |
| 4 | |
| 5 This file is part of GCC. | |
| 6 | |
| 7 GCC is free software; you can redistribute it and/or modify it under | |
| 8 the terms of the GNU General Public License as published by the Free | |
| 9 Software Foundation; either version 3, or (at your option) any later | |
| 10 version. | |
| 11 | |
| 12 GCC is distributed in the hope that it will be useful, but WITHOUT ANY | |
| 13 WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
| 14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
| 15 for more details. | |
| 16 | |
| 17 You should have received a copy of the GNU General Public License | |
| 18 along with GCC; see the file COPYING3. If not see | |
| 19 <http://www.gnu.org/licenses/>. */ | |
| 20 | |
| 21 #include "config.h" | |
| 22 #include "system.h" | |
| 23 #include "coretypes.h" | |
| 24 #include "tm.h" | |
| 25 #include "tree.h" | |
| 26 #include "rtl.h" | |
| 27 #include "tm_p.h" | |
| 28 #include "toplev.h" | |
| 29 #include "flags.h" | |
| 30 #include "ggc.h" | |
| 31 #include "timevar.h" | |
| 32 #include "params.h" | |
| 33 #include "tree-flow.h" | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
34 #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
|
35 #include "plugin.h" |
| 0 | 36 |
| 37 /* Prefer MAP_ANON(YMOUS) to /dev/zero, since we don't need to keep a | |
| 38 file open. Prefer either to valloc. */ | |
| 39 #ifdef HAVE_MMAP_ANON | |
| 40 # undef HAVE_MMAP_DEV_ZERO | |
| 41 | |
| 42 # include <sys/mman.h> | |
| 43 # ifndef MAP_FAILED | |
| 44 # define MAP_FAILED -1 | |
| 45 # endif | |
| 46 # if !defined (MAP_ANONYMOUS) && defined (MAP_ANON) | |
| 47 # define MAP_ANONYMOUS MAP_ANON | |
| 48 # endif | |
| 49 # define USING_MMAP | |
| 50 | |
| 51 #endif | |
| 52 | |
| 53 #ifdef HAVE_MMAP_DEV_ZERO | |
| 54 | |
| 55 # include <sys/mman.h> | |
| 56 # ifndef MAP_FAILED | |
| 57 # define MAP_FAILED -1 | |
| 58 # endif | |
| 59 # define USING_MMAP | |
| 60 | |
| 61 #endif | |
| 62 | |
| 63 #ifndef USING_MMAP | |
| 64 #define USING_MALLOC_PAGE_GROUPS | |
| 65 #endif | |
| 66 | |
| 67 /* Strategy: | |
| 68 | |
| 69 This garbage-collecting allocator allocates objects on one of a set | |
| 70 of pages. Each page can allocate objects of a single size only; | |
| 71 available sizes are powers of two starting at four bytes. The size | |
| 72 of an allocation request is rounded up to the next power of two | |
| 73 (`order'), and satisfied from the appropriate page. | |
| 74 | |
| 75 Each page is recorded in a page-entry, which also maintains an | |
| 76 in-use bitmap of object positions on the page. This allows the | |
| 77 allocation state of a particular object to be flipped without | |
| 78 touching the page itself. | |
| 79 | |
| 80 Each page-entry also has a context depth, which is used to track | |
| 81 pushing and popping of allocation contexts. Only objects allocated | |
| 82 in the current (highest-numbered) context may be collected. | |
| 83 | |
| 84 Page entries are arranged in an array of singly-linked lists. The | |
| 85 array is indexed by the allocation size, in bits, of the pages on | |
| 86 it; i.e. all pages on a list allocate objects of the same size. | |
| 87 Pages are ordered on the list such that all non-full pages precede | |
| 88 all full pages, with non-full pages arranged in order of decreasing | |
| 89 context depth. | |
| 90 | |
| 91 Empty pages (of all orders) are kept on a single page cache list, | |
| 92 and are considered first when new pages are required; they are | |
| 93 deallocated at the start of the next collection if they haven't | |
| 94 been recycled by then. */ | |
| 95 | |
| 96 /* Define GGC_DEBUG_LEVEL to print debugging information. | |
| 97 0: No debugging output. | |
| 98 1: GC statistics only. | |
| 99 2: Page-entry allocations/deallocations as well. | |
| 100 3: Object allocations as well. | |
| 101 4: Object marks as well. */ | |
| 102 #define GGC_DEBUG_LEVEL (0) | |
| 103 | |
| 104 #ifndef HOST_BITS_PER_PTR | |
| 105 #define HOST_BITS_PER_PTR HOST_BITS_PER_LONG | |
| 106 #endif | |
| 107 | |
| 108 | |
| 109 /* A two-level tree is used to look up the page-entry for a given | |
| 110 pointer. Two chunks of the pointer's bits are extracted to index | |
| 111 the first and second levels of the tree, as follows: | |
| 112 | |
| 113 HOST_PAGE_SIZE_BITS | |
| 114 32 | | | |
| 115 msb +----------------+----+------+------+ lsb | |
| 116 | | | | |
| 117 PAGE_L1_BITS | | |
| 118 | | | |
| 119 PAGE_L2_BITS | |
| 120 | |
| 121 The bottommost HOST_PAGE_SIZE_BITS are ignored, since page-entry | |
| 122 pages are aligned on system page boundaries. The next most | |
| 123 significant PAGE_L2_BITS and PAGE_L1_BITS are the second and first | |
| 124 index values in the lookup table, respectively. | |
| 125 | |
| 126 For 32-bit architectures and the settings below, there are no | |
| 127 leftover bits. For architectures with wider pointers, the lookup | |
| 128 tree points to a list of pages, which must be scanned to find the | |
| 129 correct one. */ | |
| 130 | |
| 131 #define PAGE_L1_BITS (8) | |
| 132 #define PAGE_L2_BITS (32 - PAGE_L1_BITS - G.lg_pagesize) | |
| 133 #define PAGE_L1_SIZE ((size_t) 1 << PAGE_L1_BITS) | |
| 134 #define PAGE_L2_SIZE ((size_t) 1 << PAGE_L2_BITS) | |
| 135 | |
| 136 #define LOOKUP_L1(p) \ | |
| 137 (((size_t) (p) >> (32 - PAGE_L1_BITS)) & ((1 << PAGE_L1_BITS) - 1)) | |
| 138 | |
| 139 #define LOOKUP_L2(p) \ | |
| 140 (((size_t) (p) >> G.lg_pagesize) & ((1 << PAGE_L2_BITS) - 1)) | |
| 141 | |
| 142 /* The number of objects per allocation page, for objects on a page of | |
| 143 the indicated ORDER. */ | |
| 144 #define OBJECTS_PER_PAGE(ORDER) objects_per_page_table[ORDER] | |
| 145 | |
| 146 /* The number of objects in P. */ | |
| 147 #define OBJECTS_IN_PAGE(P) ((P)->bytes / OBJECT_SIZE ((P)->order)) | |
| 148 | |
| 149 /* The size of an object on a page of the indicated ORDER. */ | |
| 150 #define OBJECT_SIZE(ORDER) object_size_table[ORDER] | |
| 151 | |
| 152 /* For speed, we avoid doing a general integer divide to locate the | |
| 153 offset in the allocation bitmap, by precalculating numbers M, S | |
| 154 such that (O * M) >> S == O / Z (modulo 2^32), for any offset O | |
| 155 within the page which is evenly divisible by the object size Z. */ | |
| 156 #define DIV_MULT(ORDER) inverse_table[ORDER].mult | |
| 157 #define DIV_SHIFT(ORDER) inverse_table[ORDER].shift | |
| 158 #define OFFSET_TO_BIT(OFFSET, ORDER) \ | |
| 159 (((OFFSET) * DIV_MULT (ORDER)) >> DIV_SHIFT (ORDER)) | |
| 160 | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
161 /* 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
|
162 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
|
163 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
|
164 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
|
165 |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
166 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
|
167 char c; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
168 union { |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
169 HOST_WIDEST_INT i; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
170 void *p; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
171 } u; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
172 }; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
173 |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
174 /* 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
|
175 |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
176 #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
|
177 |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
178 |
| 0 | 179 /* The number of extra orders, not corresponding to power-of-two sized |
| 180 objects. */ | |
| 181 | |
| 182 #define NUM_EXTRA_ORDERS ARRAY_SIZE (extra_order_size_table) | |
| 183 | |
| 184 #define RTL_SIZE(NSLOTS) \ | |
| 185 (RTX_HDR_SIZE + (NSLOTS) * sizeof (rtunion)) | |
| 186 | |
| 187 #define TREE_EXP_SIZE(OPS) \ | |
| 188 (sizeof (struct tree_exp) + ((OPS) - 1) * sizeof (tree)) | |
| 189 | |
| 190 /* The Ith entry is the maximum size of an object to be stored in the | |
| 191 Ith extra order. Adding a new entry to this array is the *only* | |
| 192 thing you need to do to add a new special allocation size. */ | |
| 193 | |
| 194 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
|
195 /* 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
|
196 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
|
197 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
|
198 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
|
199 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
|
200 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
|
201 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
|
202 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
|
203 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
|
204 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
|
205 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
|
206 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
|
207 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
|
208 MAX_ALIGNMENT * 15, |
| 0 | 209 sizeof (struct tree_decl_non_common), |
| 210 sizeof (struct tree_field_decl), | |
| 211 sizeof (struct tree_parm_decl), | |
| 212 sizeof (struct tree_var_decl), | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
213 sizeof (struct tree_type), |
| 0 | 214 sizeof (struct function), |
| 215 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
|
216 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
|
217 sizeof (struct loop), |
| 0 | 218 }; |
| 219 | |
| 220 /* The total number of orders. */ | |
| 221 | |
| 222 #define NUM_ORDERS (HOST_BITS_PER_PTR + NUM_EXTRA_ORDERS) | |
| 223 | |
| 224 /* Compute the smallest nonnegative number which when added to X gives | |
| 225 a multiple of F. */ | |
| 226 | |
| 227 #define ROUND_UP_VALUE(x, f) ((f) - 1 - ((f) - 1 + (x)) % (f)) | |
| 228 | |
| 229 /* Compute the smallest multiple of F that is >= X. */ | |
| 230 | |
| 231 #define ROUND_UP(x, f) (CEIL (x, f) * (f)) | |
| 232 | |
| 233 /* The Ith entry is the number of objects on a page or order I. */ | |
| 234 | |
| 235 static unsigned objects_per_page_table[NUM_ORDERS]; | |
| 236 | |
| 237 /* The Ith entry is the size of an object on a page of order I. */ | |
| 238 | |
| 239 static size_t object_size_table[NUM_ORDERS]; | |
| 240 | |
| 241 /* The Ith entry is a pair of numbers (mult, shift) such that | |
| 242 ((k * mult) >> shift) mod 2^32 == (k / OBJECT_SIZE(I)) mod 2^32, | |
| 243 for all k evenly divisible by OBJECT_SIZE(I). */ | |
| 244 | |
| 245 static struct | |
| 246 { | |
| 247 size_t mult; | |
| 248 unsigned int shift; | |
| 249 } | |
| 250 inverse_table[NUM_ORDERS]; | |
| 251 | |
| 252 /* A page_entry records the status of an allocation page. This | |
| 253 structure is dynamically sized to fit the bitmap in_use_p. */ | |
| 254 typedef struct page_entry | |
| 255 { | |
| 256 /* The next page-entry with objects of the same size, or NULL if | |
| 257 this is the last page-entry. */ | |
| 258 struct page_entry *next; | |
| 259 | |
| 260 /* The previous page-entry with objects of the same size, or NULL if | |
| 261 this is the first page-entry. The PREV pointer exists solely to | |
| 262 keep the cost of ggc_free manageable. */ | |
| 263 struct page_entry *prev; | |
| 264 | |
| 265 /* The number of bytes allocated. (This will always be a multiple | |
| 266 of the host system page size.) */ | |
| 267 size_t bytes; | |
| 268 | |
| 269 /* The address at which the memory is allocated. */ | |
| 270 char *page; | |
| 271 | |
| 272 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 273 /* Back pointer to the page group this page came from. */ | |
| 274 struct page_group *group; | |
| 275 #endif | |
| 276 | |
| 277 /* This is the index in the by_depth varray where this page table | |
| 278 can be found. */ | |
| 279 unsigned long index_by_depth; | |
| 280 | |
| 281 /* Context depth of this page. */ | |
| 282 unsigned short context_depth; | |
| 283 | |
| 284 /* The number of free objects remaining on this page. */ | |
| 285 unsigned short num_free_objects; | |
| 286 | |
| 287 /* A likely candidate for the bit position of a free object for the | |
| 288 next allocation from this page. */ | |
| 289 unsigned short next_bit_hint; | |
| 290 | |
| 291 /* The lg of size of objects allocated from this page. */ | |
| 292 unsigned char order; | |
| 293 | |
| 294 /* A bit vector indicating whether or not objects are in use. The | |
| 295 Nth bit is one if the Nth object on this page is allocated. This | |
| 296 array is dynamically sized. */ | |
| 297 unsigned long in_use_p[1]; | |
| 298 } page_entry; | |
| 299 | |
| 300 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 301 /* A page_group describes a large allocation from malloc, from which | |
| 302 we parcel out aligned pages. */ | |
| 303 typedef struct page_group | |
| 304 { | |
| 305 /* A linked list of all extant page groups. */ | |
| 306 struct page_group *next; | |
| 307 | |
| 308 /* The address we received from malloc. */ | |
| 309 char *allocation; | |
| 310 | |
| 311 /* The size of the block. */ | |
| 312 size_t alloc_size; | |
| 313 | |
| 314 /* A bitmask of pages in use. */ | |
| 315 unsigned int in_use; | |
| 316 } page_group; | |
| 317 #endif | |
| 318 | |
| 319 #if HOST_BITS_PER_PTR <= 32 | |
| 320 | |
| 321 /* On 32-bit hosts, we use a two level page table, as pictured above. */ | |
| 322 typedef page_entry **page_table[PAGE_L1_SIZE]; | |
| 323 | |
| 324 #else | |
| 325 | |
| 326 /* On 64-bit hosts, we use the same two level page tables plus a linked | |
| 327 list that disambiguates the top 32-bits. There will almost always be | |
| 328 exactly one entry in the list. */ | |
| 329 typedef struct page_table_chain | |
| 330 { | |
| 331 struct page_table_chain *next; | |
| 332 size_t high_bits; | |
| 333 page_entry **table[PAGE_L1_SIZE]; | |
| 334 } *page_table; | |
| 335 | |
| 336 #endif | |
| 337 | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
338 #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
|
339 /* 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
|
340 next collection. */ |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
341 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
|
342 { |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
343 void *object; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
344 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
|
345 }; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
346 #endif |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
347 |
| 0 | 348 /* The rest of the global variables. */ |
| 349 static struct globals | |
| 350 { | |
| 351 /* The Nth element in this array is a page with objects of size 2^N. | |
| 352 If there are any pages with free objects, they will be at the | |
| 353 head of the list. NULL if there are no page-entries for this | |
| 354 object size. */ | |
| 355 page_entry *pages[NUM_ORDERS]; | |
| 356 | |
| 357 /* The Nth element in this array is the last page with objects of | |
| 358 size 2^N. NULL if there are no page-entries for this object | |
| 359 size. */ | |
| 360 page_entry *page_tails[NUM_ORDERS]; | |
| 361 | |
| 362 /* Lookup table for associating allocation pages with object addresses. */ | |
| 363 page_table lookup; | |
| 364 | |
| 365 /* The system's page size. */ | |
| 366 size_t pagesize; | |
| 367 size_t lg_pagesize; | |
| 368 | |
| 369 /* Bytes currently allocated. */ | |
| 370 size_t allocated; | |
| 371 | |
| 372 /* Bytes currently allocated at the end of the last collection. */ | |
| 373 size_t allocated_last_gc; | |
| 374 | |
| 375 /* Total amount of memory mapped. */ | |
| 376 size_t bytes_mapped; | |
| 377 | |
| 378 /* Bit N set if any allocations have been done at context depth N. */ | |
| 379 unsigned long context_depth_allocations; | |
| 380 | |
| 381 /* Bit N set if any collections have been done at context depth N. */ | |
| 382 unsigned long context_depth_collections; | |
| 383 | |
| 384 /* The current depth in the context stack. */ | |
| 385 unsigned short context_depth; | |
| 386 | |
| 387 /* A file descriptor open to /dev/zero for reading. */ | |
| 388 #if defined (HAVE_MMAP_DEV_ZERO) | |
| 389 int dev_zero_fd; | |
| 390 #endif | |
| 391 | |
| 392 /* A cache of free system pages. */ | |
| 393 page_entry *free_pages; | |
| 394 | |
| 395 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 396 page_group *page_groups; | |
| 397 #endif | |
| 398 | |
| 399 /* The file descriptor for debugging output. */ | |
| 400 FILE *debug_file; | |
| 401 | |
| 402 /* Current number of elements in use in depth below. */ | |
| 403 unsigned int depth_in_use; | |
| 404 | |
| 405 /* Maximum number of elements that can be used before resizing. */ | |
| 406 unsigned int depth_max; | |
| 407 | |
| 408 /* Each element of this array is an index in by_depth where the given | |
| 409 depth starts. This structure is indexed by that given depth we | |
| 410 are interested in. */ | |
| 411 unsigned int *depth; | |
| 412 | |
| 413 /* Current number of elements in use in by_depth below. */ | |
| 414 unsigned int by_depth_in_use; | |
| 415 | |
| 416 /* Maximum number of elements that can be used before resizing. */ | |
| 417 unsigned int by_depth_max; | |
| 418 | |
| 419 /* Each element of this array is a pointer to a page_entry, all | |
| 420 page_entries can be found in here by increasing depth. | |
| 421 index_by_depth in the page_entry is the index into this data | |
| 422 structure where that page_entry can be found. This is used to | |
| 423 speed up finding all page_entries at a particular depth. */ | |
| 424 page_entry **by_depth; | |
| 425 | |
| 426 /* Each element is a pointer to the saved in_use_p bits, if any, | |
| 427 zero otherwise. We allocate them all together, to enable a | |
| 428 better runtime data access pattern. */ | |
| 429 unsigned long **save_in_use; | |
| 430 | |
| 431 #ifdef ENABLE_GC_ALWAYS_COLLECT | |
| 432 /* List of free objects to be verified as actually free on the | |
| 433 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
|
434 struct free_object *free_object_list; |
| 0 | 435 #endif |
| 436 | |
| 437 #ifdef GATHER_STATISTICS | |
| 438 struct | |
| 439 { | |
| 440 /* Total memory allocated with ggc_alloc. */ | |
| 441 unsigned long long total_allocated; | |
| 442 /* Total overhead for memory to be allocated with ggc_alloc. */ | |
| 443 unsigned long long total_overhead; | |
| 444 | |
| 445 /* Total allocations and overhead for sizes less than 32, 64 and 128. | |
| 446 These sizes are interesting because they are typical cache line | |
| 447 sizes. */ | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
448 |
| 0 | 449 unsigned long long total_allocated_under32; |
| 450 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
|
451 |
| 0 | 452 unsigned long long total_allocated_under64; |
| 453 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
|
454 |
| 0 | 455 unsigned long long total_allocated_under128; |
| 456 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
|
457 |
| 0 | 458 /* The allocations for each of the allocation orders. */ |
| 459 unsigned long long total_allocated_per_order[NUM_ORDERS]; | |
| 460 | |
| 461 /* The overhead for each of the allocation orders. */ | |
| 462 unsigned long long total_overhead_per_order[NUM_ORDERS]; | |
| 463 } stats; | |
| 464 #endif | |
| 465 } G; | |
| 466 | |
| 467 /* The size in bytes required to maintain a bitmap for the objects | |
| 468 on a page-entry. */ | |
| 469 #define BITMAP_SIZE(Num_objects) \ | |
| 470 (CEIL ((Num_objects), HOST_BITS_PER_LONG) * sizeof(long)) | |
| 471 | |
| 472 /* Allocate pages in chunks of this size, to throttle calls to memory | |
| 473 allocation routines. The first page is used, the rest go onto the | |
| 474 free list. This cannot be larger than HOST_BITS_PER_INT for the | |
| 475 in_use bitmask for page_group. Hosts that need a different value | |
| 476 can override this by defining GGC_QUIRE_SIZE explicitly. */ | |
| 477 #ifndef GGC_QUIRE_SIZE | |
| 478 # ifdef USING_MMAP | |
| 479 # define GGC_QUIRE_SIZE 256 | |
| 480 # else | |
| 481 # define GGC_QUIRE_SIZE 16 | |
| 482 # endif | |
| 483 #endif | |
| 484 | |
| 485 /* Initial guess as to how many page table entries we might need. */ | |
| 486 #define INITIAL_PTE_COUNT 128 | |
| 487 | |
| 488 static int ggc_allocated_p (const void *); | |
| 489 static page_entry *lookup_page_table_entry (const void *); | |
| 490 static void set_page_table_entry (void *, page_entry *); | |
| 491 #ifdef USING_MMAP | |
| 492 static char *alloc_anon (char *, size_t); | |
| 493 #endif | |
| 494 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 495 static size_t page_group_index (char *, char *); | |
| 496 static void set_page_group_in_use (page_group *, char *); | |
| 497 static void clear_page_group_in_use (page_group *, char *); | |
| 498 #endif | |
| 499 static struct page_entry * alloc_page (unsigned); | |
| 500 static void free_page (struct page_entry *); | |
| 501 static void release_pages (void); | |
| 502 static void clear_marks (void); | |
| 503 static void sweep_pages (void); | |
| 504 static void ggc_recalculate_in_use_p (page_entry *); | |
| 505 static void compute_inverse (unsigned); | |
| 506 static inline void adjust_depth (void); | |
| 507 static void move_ptes_to_front (int, int); | |
| 508 | |
| 509 void debug_print_page_list (int); | |
| 510 static void push_depth (unsigned int); | |
| 511 static void push_by_depth (page_entry *, unsigned long *); | |
| 512 | |
| 513 /* Push an entry onto G.depth. */ | |
| 514 | |
| 515 inline static void | |
| 516 push_depth (unsigned int i) | |
| 517 { | |
| 518 if (G.depth_in_use >= G.depth_max) | |
| 519 { | |
| 520 G.depth_max *= 2; | |
| 521 G.depth = XRESIZEVEC (unsigned int, G.depth, G.depth_max); | |
| 522 } | |
| 523 G.depth[G.depth_in_use++] = i; | |
| 524 } | |
| 525 | |
| 526 /* Push an entry onto G.by_depth and G.save_in_use. */ | |
| 527 | |
| 528 inline static void | |
| 529 push_by_depth (page_entry *p, unsigned long *s) | |
| 530 { | |
| 531 if (G.by_depth_in_use >= G.by_depth_max) | |
| 532 { | |
| 533 G.by_depth_max *= 2; | |
| 534 G.by_depth = XRESIZEVEC (page_entry *, G.by_depth, G.by_depth_max); | |
| 535 G.save_in_use = XRESIZEVEC (unsigned long *, G.save_in_use, | |
| 536 G.by_depth_max); | |
| 537 } | |
| 538 G.by_depth[G.by_depth_in_use] = p; | |
| 539 G.save_in_use[G.by_depth_in_use++] = s; | |
| 540 } | |
| 541 | |
| 542 #if (GCC_VERSION < 3001) | |
| 543 #define prefetch(X) ((void) X) | |
| 544 #else | |
| 545 #define prefetch(X) __builtin_prefetch (X) | |
| 546 #endif | |
| 547 | |
| 548 #define save_in_use_p_i(__i) \ | |
| 549 (G.save_in_use[__i]) | |
| 550 #define save_in_use_p(__p) \ | |
| 551 (save_in_use_p_i (__p->index_by_depth)) | |
| 552 | |
| 553 /* Returns nonzero if P was allocated in GC'able memory. */ | |
| 554 | |
| 555 static inline int | |
| 556 ggc_allocated_p (const void *p) | |
| 557 { | |
| 558 page_entry ***base; | |
| 559 size_t L1, L2; | |
| 560 | |
| 561 #if HOST_BITS_PER_PTR <= 32 | |
| 562 base = &G.lookup[0]; | |
| 563 #else | |
| 564 page_table table = G.lookup; | |
| 565 size_t high_bits = (size_t) p & ~ (size_t) 0xffffffff; | |
| 566 while (1) | |
| 567 { | |
| 568 if (table == NULL) | |
| 569 return 0; | |
| 570 if (table->high_bits == high_bits) | |
| 571 break; | |
| 572 table = table->next; | |
| 573 } | |
| 574 base = &table->table[0]; | |
| 575 #endif | |
| 576 | |
| 577 /* Extract the level 1 and 2 indices. */ | |
| 578 L1 = LOOKUP_L1 (p); | |
| 579 L2 = LOOKUP_L2 (p); | |
| 580 | |
| 581 return base[L1] && base[L1][L2]; | |
| 582 } | |
| 583 | |
| 584 /* Traverse the page table and find the entry for a page. | |
| 585 Die (probably) if the object wasn't allocated via GC. */ | |
| 586 | |
| 587 static inline page_entry * | |
| 588 lookup_page_table_entry (const void *p) | |
| 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; | |
| 597 size_t high_bits = (size_t) p & ~ (size_t) 0xffffffff; | |
| 598 while (table->high_bits != high_bits) | |
| 599 table = table->next; | |
| 600 base = &table->table[0]; | |
| 601 #endif | |
| 602 | |
| 603 /* Extract the level 1 and 2 indices. */ | |
| 604 L1 = LOOKUP_L1 (p); | |
| 605 L2 = LOOKUP_L2 (p); | |
| 606 | |
| 607 return base[L1][L2]; | |
| 608 } | |
| 609 | |
| 610 /* Set the page table entry for a page. */ | |
| 611 | |
| 612 static void | |
| 613 set_page_table_entry (void *p, page_entry *entry) | |
| 614 { | |
| 615 page_entry ***base; | |
| 616 size_t L1, L2; | |
| 617 | |
| 618 #if HOST_BITS_PER_PTR <= 32 | |
| 619 base = &G.lookup[0]; | |
| 620 #else | |
| 621 page_table table; | |
| 622 size_t high_bits = (size_t) p & ~ (size_t) 0xffffffff; | |
| 623 for (table = G.lookup; table; table = table->next) | |
| 624 if (table->high_bits == high_bits) | |
| 625 goto found; | |
| 626 | |
| 627 /* Not found -- allocate a new table. */ | |
| 628 table = XCNEW (struct page_table_chain); | |
| 629 table->next = G.lookup; | |
| 630 table->high_bits = high_bits; | |
| 631 G.lookup = table; | |
| 632 found: | |
| 633 base = &table->table[0]; | |
| 634 #endif | |
| 635 | |
| 636 /* Extract the level 1 and 2 indices. */ | |
| 637 L1 = LOOKUP_L1 (p); | |
| 638 L2 = LOOKUP_L2 (p); | |
| 639 | |
| 640 if (base[L1] == NULL) | |
| 641 base[L1] = XCNEWVEC (page_entry *, PAGE_L2_SIZE); | |
| 642 | |
| 643 base[L1][L2] = entry; | |
| 644 } | |
| 645 | |
| 646 /* Prints the page-entry for object size ORDER, for debugging. */ | |
| 647 | |
| 648 void | |
| 649 debug_print_page_list (int order) | |
| 650 { | |
| 651 page_entry *p; | |
| 652 printf ("Head=%p, Tail=%p:\n", (void *) G.pages[order], | |
| 653 (void *) G.page_tails[order]); | |
| 654 p = G.pages[order]; | |
| 655 while (p != NULL) | |
| 656 { | |
| 657 printf ("%p(%1d|%3d) -> ", (void *) p, p->context_depth, | |
| 658 p->num_free_objects); | |
| 659 p = p->next; | |
| 660 } | |
| 661 printf ("NULL\n"); | |
| 662 fflush (stdout); | |
| 663 } | |
| 664 | |
| 665 #ifdef USING_MMAP | |
| 666 /* Allocate SIZE bytes of anonymous memory, preferably near PREF, | |
| 667 (if non-null). The ifdef structure here is intended to cause a | |
| 668 compile error unless exactly one of the HAVE_* is defined. */ | |
| 669 | |
| 670 static inline char * | |
| 671 alloc_anon (char *pref ATTRIBUTE_UNUSED, size_t size) | |
| 672 { | |
| 673 #ifdef HAVE_MMAP_ANON | |
| 674 char *page = (char *) mmap (pref, size, PROT_READ | PROT_WRITE, | |
| 675 MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); | |
| 676 #endif | |
| 677 #ifdef HAVE_MMAP_DEV_ZERO | |
| 678 char *page = (char *) mmap (pref, size, PROT_READ | PROT_WRITE, | |
| 679 MAP_PRIVATE, G.dev_zero_fd, 0); | |
| 680 #endif | |
| 681 | |
| 682 if (page == (char *) MAP_FAILED) | |
| 683 { | |
| 684 perror ("virtual memory exhausted"); | |
| 685 exit (FATAL_EXIT_CODE); | |
| 686 } | |
| 687 | |
| 688 /* Remember that we allocated this memory. */ | |
| 689 G.bytes_mapped += size; | |
| 690 | |
| 691 /* Pretend we don't have access to the allocated pages. We'll enable | |
| 692 access to smaller pieces of the area in ggc_alloc. Discard the | |
| 693 handle to avoid handle leak. */ | |
| 694 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (page, size)); | |
| 695 | |
| 696 return page; | |
| 697 } | |
| 698 #endif | |
| 699 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 700 /* Compute the index for this page into the page group. */ | |
| 701 | |
| 702 static inline size_t | |
| 703 page_group_index (char *allocation, char *page) | |
| 704 { | |
| 705 return (size_t) (page - allocation) >> G.lg_pagesize; | |
| 706 } | |
| 707 | |
| 708 /* Set and clear the in_use bit for this page in the page group. */ | |
| 709 | |
| 710 static inline void | |
| 711 set_page_group_in_use (page_group *group, char *page) | |
| 712 { | |
| 713 group->in_use |= 1 << page_group_index (group->allocation, page); | |
| 714 } | |
| 715 | |
| 716 static inline void | |
| 717 clear_page_group_in_use (page_group *group, char *page) | |
| 718 { | |
| 719 group->in_use &= ~(1 << page_group_index (group->allocation, page)); | |
| 720 } | |
| 721 #endif | |
| 722 | |
| 723 /* Allocate a new page for allocating objects of size 2^ORDER, | |
| 724 and return an entry for it. The entry is not added to the | |
| 725 appropriate page_table list. */ | |
| 726 | |
| 727 static inline struct page_entry * | |
| 728 alloc_page (unsigned order) | |
| 729 { | |
| 730 struct page_entry *entry, *p, **pp; | |
| 731 char *page; | |
| 732 size_t num_objects; | |
| 733 size_t bitmap_size; | |
| 734 size_t page_entry_size; | |
| 735 size_t entry_size; | |
| 736 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 737 page_group *group; | |
| 738 #endif | |
| 739 | |
| 740 num_objects = OBJECTS_PER_PAGE (order); | |
| 741 bitmap_size = BITMAP_SIZE (num_objects + 1); | |
| 742 page_entry_size = sizeof (page_entry) - sizeof (long) + bitmap_size; | |
| 743 entry_size = num_objects * OBJECT_SIZE (order); | |
| 744 if (entry_size < G.pagesize) | |
| 745 entry_size = G.pagesize; | |
| 746 | |
| 747 entry = NULL; | |
| 748 page = NULL; | |
| 749 | |
| 750 /* Check the list of free pages for one we can use. */ | |
| 751 for (pp = &G.free_pages, p = *pp; p; pp = &p->next, p = *pp) | |
| 752 if (p->bytes == entry_size) | |
| 753 break; | |
| 754 | |
| 755 if (p != NULL) | |
| 756 { | |
| 757 /* Recycle the allocated memory from this page ... */ | |
| 758 *pp = p->next; | |
| 759 page = p->page; | |
| 760 | |
| 761 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 762 group = p->group; | |
| 763 #endif | |
| 764 | |
| 765 /* ... and, if possible, the page entry itself. */ | |
| 766 if (p->order == order) | |
| 767 { | |
| 768 entry = p; | |
| 769 memset (entry, 0, page_entry_size); | |
| 770 } | |
| 771 else | |
| 772 free (p); | |
| 773 } | |
| 774 #ifdef USING_MMAP | |
| 775 else if (entry_size == G.pagesize) | |
| 776 { | |
| 777 /* We want just one page. Allocate a bunch of them and put the | |
| 778 extras on the freelist. (Can only do this optimization with | |
| 779 mmap for backing store.) */ | |
| 780 struct page_entry *e, *f = G.free_pages; | |
| 781 int i; | |
| 782 | |
| 783 page = alloc_anon (NULL, G.pagesize * GGC_QUIRE_SIZE); | |
| 784 | |
| 785 /* This loop counts down so that the chain will be in ascending | |
| 786 memory order. */ | |
| 787 for (i = GGC_QUIRE_SIZE - 1; i >= 1; i--) | |
| 788 { | |
| 789 e = XCNEWVAR (struct page_entry, page_entry_size); | |
| 790 e->order = order; | |
| 791 e->bytes = G.pagesize; | |
| 792 e->page = page + (i << G.lg_pagesize); | |
| 793 e->next = f; | |
| 794 f = e; | |
| 795 } | |
| 796 | |
| 797 G.free_pages = f; | |
| 798 } | |
| 799 else | |
| 800 page = alloc_anon (NULL, entry_size); | |
| 801 #endif | |
| 802 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 803 else | |
| 804 { | |
| 805 /* Allocate a large block of memory and serve out the aligned | |
| 806 pages therein. This results in much less memory wastage | |
| 807 than the traditional implementation of valloc. */ | |
| 808 | |
| 809 char *allocation, *a, *enda; | |
| 810 size_t alloc_size, head_slop, tail_slop; | |
| 811 int multiple_pages = (entry_size == G.pagesize); | |
| 812 | |
| 813 if (multiple_pages) | |
| 814 alloc_size = GGC_QUIRE_SIZE * G.pagesize; | |
| 815 else | |
| 816 alloc_size = entry_size + G.pagesize - 1; | |
| 817 allocation = XNEWVEC (char, alloc_size); | |
| 818 | |
| 819 page = (char *) (((size_t) allocation + G.pagesize - 1) & -G.pagesize); | |
| 820 head_slop = page - allocation; | |
| 821 if (multiple_pages) | |
| 822 tail_slop = ((size_t) allocation + alloc_size) & (G.pagesize - 1); | |
| 823 else | |
| 824 tail_slop = alloc_size - entry_size - head_slop; | |
| 825 enda = allocation + alloc_size - tail_slop; | |
| 826 | |
| 827 /* We allocated N pages, which are likely not aligned, leaving | |
| 828 us with N-1 usable pages. We plan to place the page_group | |
| 829 structure somewhere in the slop. */ | |
| 830 if (head_slop >= sizeof (page_group)) | |
| 831 group = (page_group *)page - 1; | |
| 832 else | |
| 833 { | |
| 834 /* We magically got an aligned allocation. Too bad, we have | |
| 835 to waste a page anyway. */ | |
| 836 if (tail_slop == 0) | |
| 837 { | |
| 838 enda -= G.pagesize; | |
| 839 tail_slop += G.pagesize; | |
| 840 } | |
| 841 gcc_assert (tail_slop >= sizeof (page_group)); | |
| 842 group = (page_group *)enda; | |
| 843 tail_slop -= sizeof (page_group); | |
| 844 } | |
| 845 | |
| 846 /* Remember that we allocated this memory. */ | |
| 847 group->next = G.page_groups; | |
| 848 group->allocation = allocation; | |
| 849 group->alloc_size = alloc_size; | |
| 850 group->in_use = 0; | |
| 851 G.page_groups = group; | |
| 852 G.bytes_mapped += alloc_size; | |
| 853 | |
| 854 /* If we allocated multiple pages, put the rest on the free list. */ | |
| 855 if (multiple_pages) | |
| 856 { | |
| 857 struct page_entry *e, *f = G.free_pages; | |
| 858 for (a = enda - G.pagesize; a != page; a -= G.pagesize) | |
| 859 { | |
| 860 e = XCNEWVAR (struct page_entry, page_entry_size); | |
| 861 e->order = order; | |
| 862 e->bytes = G.pagesize; | |
| 863 e->page = a; | |
| 864 e->group = group; | |
| 865 e->next = f; | |
| 866 f = e; | |
| 867 } | |
| 868 G.free_pages = f; | |
| 869 } | |
| 870 } | |
| 871 #endif | |
| 872 | |
| 873 if (entry == NULL) | |
| 874 entry = XCNEWVAR (struct page_entry, page_entry_size); | |
| 875 | |
| 876 entry->bytes = entry_size; | |
| 877 entry->page = page; | |
| 878 entry->context_depth = G.context_depth; | |
| 879 entry->order = order; | |
| 880 entry->num_free_objects = num_objects; | |
| 881 entry->next_bit_hint = 1; | |
| 882 | |
| 883 G.context_depth_allocations |= (unsigned long)1 << G.context_depth; | |
| 884 | |
| 885 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 886 entry->group = group; | |
| 887 set_page_group_in_use (group, page); | |
| 888 #endif | |
| 889 | |
| 890 /* Set the one-past-the-end in-use bit. This acts as a sentry as we | |
| 891 increment the hint. */ | |
| 892 entry->in_use_p[num_objects / HOST_BITS_PER_LONG] | |
| 893 = (unsigned long) 1 << (num_objects % HOST_BITS_PER_LONG); | |
| 894 | |
| 895 set_page_table_entry (page, entry); | |
| 896 | |
| 897 if (GGC_DEBUG_LEVEL >= 2) | |
| 898 fprintf (G.debug_file, | |
| 899 "Allocating page at %p, object size=%lu, data %p-%p\n", | |
| 900 (void *) entry, (unsigned long) OBJECT_SIZE (order), page, | |
| 901 page + entry_size - 1); | |
| 902 | |
| 903 return entry; | |
| 904 } | |
| 905 | |
| 906 /* Adjust the size of G.depth so that no index greater than the one | |
| 907 used by the top of the G.by_depth is used. */ | |
| 908 | |
| 909 static inline void | |
| 910 adjust_depth (void) | |
| 911 { | |
| 912 page_entry *top; | |
| 913 | |
| 914 if (G.by_depth_in_use) | |
| 915 { | |
| 916 top = G.by_depth[G.by_depth_in_use-1]; | |
| 917 | |
| 918 /* Peel back indices in depth that index into by_depth, so that | |
| 919 as new elements are added to by_depth, we note the indices | |
| 920 of those elements, if they are for new context depths. */ | |
| 921 while (G.depth_in_use > (size_t)top->context_depth+1) | |
| 922 --G.depth_in_use; | |
| 923 } | |
| 924 } | |
| 925 | |
| 926 /* For a page that is no longer needed, put it on the free page list. */ | |
| 927 | |
| 928 static void | |
| 929 free_page (page_entry *entry) | |
| 930 { | |
| 931 if (GGC_DEBUG_LEVEL >= 2) | |
| 932 fprintf (G.debug_file, | |
| 933 "Deallocating page at %p, data %p-%p\n", (void *) entry, | |
| 934 entry->page, entry->page + entry->bytes - 1); | |
| 935 | |
| 936 /* Mark the page as inaccessible. Discard the handle to avoid handle | |
| 937 leak. */ | |
| 938 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (entry->page, entry->bytes)); | |
| 939 | |
| 940 set_page_table_entry (entry->page, NULL); | |
| 941 | |
| 942 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 943 clear_page_group_in_use (entry->group, entry->page); | |
| 944 #endif | |
| 945 | |
| 946 if (G.by_depth_in_use > 1) | |
| 947 { | |
| 948 page_entry *top = G.by_depth[G.by_depth_in_use-1]; | |
| 949 int i = entry->index_by_depth; | |
| 950 | |
| 951 /* We cannot free a page from a context deeper than the current | |
| 952 one. */ | |
| 953 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
|
954 |
| 0 | 955 /* Put top element into freed slot. */ |
| 956 G.by_depth[i] = top; | |
| 957 G.save_in_use[i] = G.save_in_use[G.by_depth_in_use-1]; | |
| 958 top->index_by_depth = i; | |
| 959 } | |
| 960 --G.by_depth_in_use; | |
| 961 | |
| 962 adjust_depth (); | |
| 963 | |
| 964 entry->next = G.free_pages; | |
| 965 G.free_pages = entry; | |
| 966 } | |
| 967 | |
| 968 /* Release the free page cache to the system. */ | |
| 969 | |
| 970 static void | |
| 971 release_pages (void) | |
| 972 { | |
| 973 #ifdef USING_MMAP | |
| 974 page_entry *p, *next; | |
| 975 char *start; | |
| 976 size_t len; | |
| 977 | |
| 978 /* Gather up adjacent pages so they are unmapped together. */ | |
| 979 p = G.free_pages; | |
| 980 | |
| 981 while (p) | |
| 982 { | |
| 983 start = p->page; | |
| 984 next = p->next; | |
| 985 len = p->bytes; | |
| 986 free (p); | |
| 987 p = next; | |
| 988 | |
| 989 while (p && p->page == start + len) | |
| 990 { | |
| 991 next = p->next; | |
| 992 len += p->bytes; | |
| 993 free (p); | |
| 994 p = next; | |
| 995 } | |
| 996 | |
| 997 munmap (start, len); | |
| 998 G.bytes_mapped -= len; | |
| 999 } | |
| 1000 | |
| 1001 G.free_pages = NULL; | |
| 1002 #endif | |
| 1003 #ifdef USING_MALLOC_PAGE_GROUPS | |
| 1004 page_entry **pp, *p; | |
| 1005 page_group **gp, *g; | |
| 1006 | |
| 1007 /* Remove all pages from free page groups from the list. */ | |
| 1008 pp = &G.free_pages; | |
| 1009 while ((p = *pp) != NULL) | |
| 1010 if (p->group->in_use == 0) | |
| 1011 { | |
| 1012 *pp = p->next; | |
| 1013 free (p); | |
| 1014 } | |
| 1015 else | |
| 1016 pp = &p->next; | |
| 1017 | |
| 1018 /* Remove all free page groups, and release the storage. */ | |
| 1019 gp = &G.page_groups; | |
| 1020 while ((g = *gp) != NULL) | |
| 1021 if (g->in_use == 0) | |
| 1022 { | |
| 1023 *gp = g->next; | |
| 1024 G.bytes_mapped -= g->alloc_size; | |
| 1025 free (g->allocation); | |
| 1026 } | |
| 1027 else | |
| 1028 gp = &g->next; | |
| 1029 #endif | |
| 1030 } | |
| 1031 | |
| 1032 /* This table provides a fast way to determine ceil(log_2(size)) for | |
| 1033 allocation requests. The minimum allocation size is eight bytes. */ | |
| 1034 #define NUM_SIZE_LOOKUP 512 | |
| 1035 static unsigned char size_lookup[NUM_SIZE_LOOKUP] = | |
| 1036 { | |
| 1037 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, | |
| 1038 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, | |
| 1039 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, | |
| 1040 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, | |
| 1041 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
| 1042 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
| 1043 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
| 1044 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, | |
| 1045 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1046 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1047 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1048 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1049 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1050 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1051 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1052 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, | |
| 1053 8, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1054 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1055 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1056 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1057 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1058 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1059 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1060 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1061 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1062 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1063 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1064 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1065 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1066 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1067 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, | |
| 1068 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9 | |
| 1069 }; | |
| 1070 | |
| 1071 /* Typed allocation function. Does nothing special in this collector. */ | |
| 1072 | |
| 1073 void * | |
| 1074 ggc_alloc_typed_stat (enum gt_types_enum type ATTRIBUTE_UNUSED, size_t size | |
| 1075 MEM_STAT_DECL) | |
| 1076 { | |
| 1077 return ggc_alloc_stat (size PASS_MEM_STAT); | |
| 1078 } | |
| 1079 | |
| 1080 /* Allocate a chunk of memory of SIZE bytes. Its contents are undefined. */ | |
| 1081 | |
| 1082 void * | |
| 1083 ggc_alloc_stat (size_t size MEM_STAT_DECL) | |
| 1084 { | |
| 1085 size_t order, word, bit, object_offset, object_size; | |
| 1086 struct page_entry *entry; | |
| 1087 void *result; | |
| 1088 | |
| 1089 if (size < NUM_SIZE_LOOKUP) | |
| 1090 { | |
| 1091 order = size_lookup[size]; | |
| 1092 object_size = OBJECT_SIZE (order); | |
| 1093 } | |
| 1094 else | |
| 1095 { | |
| 1096 order = 10; | |
| 1097 while (size > (object_size = OBJECT_SIZE (order))) | |
| 1098 order++; | |
| 1099 } | |
| 1100 | |
| 1101 /* If there are non-full pages for this size allocation, they are at | |
| 1102 the head of the list. */ | |
| 1103 entry = G.pages[order]; | |
| 1104 | |
| 1105 /* If there is no page for this object size, or all pages in this | |
| 1106 context are full, allocate a new page. */ | |
| 1107 if (entry == NULL || entry->num_free_objects == 0) | |
| 1108 { | |
| 1109 struct page_entry *new_entry; | |
| 1110 new_entry = alloc_page (order); | |
| 1111 | |
| 1112 new_entry->index_by_depth = G.by_depth_in_use; | |
| 1113 push_by_depth (new_entry, 0); | |
| 1114 | |
| 1115 /* We can skip context depths, if we do, make sure we go all the | |
| 1116 way to the new depth. */ | |
| 1117 while (new_entry->context_depth >= G.depth_in_use) | |
| 1118 push_depth (G.by_depth_in_use-1); | |
| 1119 | |
| 1120 /* If this is the only entry, it's also the tail. If it is not | |
| 1121 the only entry, then we must update the PREV pointer of the | |
| 1122 ENTRY (G.pages[order]) to point to our new page entry. */ | |
| 1123 if (entry == NULL) | |
| 1124 G.page_tails[order] = new_entry; | |
| 1125 else | |
| 1126 entry->prev = new_entry; | |
| 1127 | |
| 1128 /* Put new pages at the head of the page list. By definition the | |
| 1129 entry at the head of the list always has a NULL pointer. */ | |
| 1130 new_entry->next = entry; | |
| 1131 new_entry->prev = NULL; | |
| 1132 entry = new_entry; | |
| 1133 G.pages[order] = new_entry; | |
| 1134 | |
| 1135 /* For a new page, we know the word and bit positions (in the | |
| 1136 in_use bitmap) of the first available object -- they're zero. */ | |
| 1137 new_entry->next_bit_hint = 1; | |
| 1138 word = 0; | |
| 1139 bit = 0; | |
| 1140 object_offset = 0; | |
| 1141 } | |
| 1142 else | |
| 1143 { | |
| 1144 /* First try to use the hint left from the previous allocation | |
| 1145 to locate a clear bit in the in-use bitmap. We've made sure | |
| 1146 that the one-past-the-end bit is always set, so if the hint | |
| 1147 has run over, this test will fail. */ | |
| 1148 unsigned hint = entry->next_bit_hint; | |
| 1149 word = hint / HOST_BITS_PER_LONG; | |
| 1150 bit = hint % HOST_BITS_PER_LONG; | |
| 1151 | |
| 1152 /* If the hint didn't work, scan the bitmap from the beginning. */ | |
| 1153 if ((entry->in_use_p[word] >> bit) & 1) | |
| 1154 { | |
| 1155 word = bit = 0; | |
| 1156 while (~entry->in_use_p[word] == 0) | |
| 1157 ++word; | |
| 1158 | |
| 1159 #if GCC_VERSION >= 3004 | |
| 1160 bit = __builtin_ctzl (~entry->in_use_p[word]); | |
| 1161 #else | |
| 1162 while ((entry->in_use_p[word] >> bit) & 1) | |
| 1163 ++bit; | |
| 1164 #endif | |
| 1165 | |
| 1166 hint = word * HOST_BITS_PER_LONG + bit; | |
| 1167 } | |
| 1168 | |
| 1169 /* Next time, try the next bit. */ | |
| 1170 entry->next_bit_hint = hint + 1; | |
| 1171 | |
| 1172 object_offset = hint * object_size; | |
| 1173 } | |
| 1174 | |
| 1175 /* Set the in-use bit. */ | |
| 1176 entry->in_use_p[word] |= ((unsigned long) 1 << bit); | |
| 1177 | |
| 1178 /* Keep a running total of the number of free objects. If this page | |
| 1179 fills up, we may have to move it to the end of the list if the | |
| 1180 next page isn't full. If the next page is full, all subsequent | |
| 1181 pages are full, so there's no need to move it. */ | |
| 1182 if (--entry->num_free_objects == 0 | |
| 1183 && entry->next != NULL | |
| 1184 && entry->next->num_free_objects > 0) | |
| 1185 { | |
| 1186 /* We have a new head for the list. */ | |
| 1187 G.pages[order] = entry->next; | |
| 1188 | |
| 1189 /* We are moving ENTRY to the end of the page table list. | |
| 1190 The new page at the head of the list will have NULL in | |
| 1191 its PREV field and ENTRY will have NULL in its NEXT field. */ | |
| 1192 entry->next->prev = NULL; | |
| 1193 entry->next = NULL; | |
| 1194 | |
| 1195 /* Append ENTRY to the tail of the list. */ | |
| 1196 entry->prev = G.page_tails[order]; | |
| 1197 G.page_tails[order]->next = entry; | |
| 1198 G.page_tails[order] = entry; | |
| 1199 } | |
| 1200 | |
| 1201 /* Calculate the object's address. */ | |
| 1202 result = entry->page + object_offset; | |
| 1203 #ifdef GATHER_STATISTICS | |
| 1204 ggc_record_overhead (OBJECT_SIZE (order), OBJECT_SIZE (order) - size, | |
| 1205 result PASS_MEM_STAT); | |
| 1206 #endif | |
| 1207 | |
| 1208 #ifdef ENABLE_GC_CHECKING | |
| 1209 /* Keep poisoning-by-writing-0xaf the object, in an attempt to keep the | |
| 1210 exact same semantics in presence of memory bugs, regardless of | |
| 1211 ENABLE_VALGRIND_CHECKING. We override this request below. Drop the | |
| 1212 handle to avoid handle leak. */ | |
| 1213 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (result, object_size)); | |
| 1214 | |
| 1215 /* `Poison' the entire allocated object, including any padding at | |
| 1216 the end. */ | |
| 1217 memset (result, 0xaf, object_size); | |
| 1218 | |
| 1219 /* Make the bytes after the end of the object unaccessible. Discard the | |
| 1220 handle to avoid handle leak. */ | |
| 1221 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS ((char *) result + size, | |
| 1222 object_size - size)); | |
| 1223 #endif | |
| 1224 | |
| 1225 /* Tell Valgrind that the memory is there, but its content isn't | |
| 1226 defined. The bytes at the end of the object are still marked | |
| 1227 unaccessible. */ | |
| 1228 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (result, size)); | |
| 1229 | |
| 1230 /* Keep track of how many bytes are being allocated. This | |
| 1231 information is used in deciding when to collect. */ | |
| 1232 G.allocated += object_size; | |
| 1233 | |
| 1234 /* For timevar statistics. */ | |
| 1235 timevar_ggc_mem_total += object_size; | |
| 1236 | |
| 1237 #ifdef GATHER_STATISTICS | |
| 1238 { | |
| 1239 size_t overhead = object_size - size; | |
| 1240 | |
| 1241 G.stats.total_overhead += overhead; | |
| 1242 G.stats.total_allocated += object_size; | |
| 1243 G.stats.total_overhead_per_order[order] += overhead; | |
| 1244 G.stats.total_allocated_per_order[order] += object_size; | |
| 1245 | |
| 1246 if (size <= 32) | |
| 1247 { | |
| 1248 G.stats.total_overhead_under32 += overhead; | |
| 1249 G.stats.total_allocated_under32 += object_size; | |
| 1250 } | |
| 1251 if (size <= 64) | |
| 1252 { | |
| 1253 G.stats.total_overhead_under64 += overhead; | |
| 1254 G.stats.total_allocated_under64 += object_size; | |
| 1255 } | |
| 1256 if (size <= 128) | |
| 1257 { | |
| 1258 G.stats.total_overhead_under128 += overhead; | |
| 1259 G.stats.total_allocated_under128 += object_size; | |
| 1260 } | |
| 1261 } | |
| 1262 #endif | |
| 1263 | |
| 1264 if (GGC_DEBUG_LEVEL >= 3) | |
| 1265 fprintf (G.debug_file, | |
| 1266 "Allocating object, requested size=%lu, actual=%lu at %p on %p\n", | |
| 1267 (unsigned long) size, (unsigned long) object_size, result, | |
| 1268 (void *) entry); | |
| 1269 | |
| 1270 return result; | |
| 1271 } | |
| 1272 | |
| 1273 /* Mark function for strings. */ | |
| 1274 | |
| 1275 void | |
| 1276 gt_ggc_m_S (const void *p) | |
| 1277 { | |
| 1278 page_entry *entry; | |
| 1279 unsigned bit, word; | |
| 1280 unsigned long mask; | |
| 1281 unsigned long offset; | |
| 1282 | |
| 1283 if (!p || !ggc_allocated_p (p)) | |
| 1284 return; | |
| 1285 | |
| 1286 /* Look up the page on which the object is alloced. . */ | |
| 1287 entry = lookup_page_table_entry (p); | |
| 1288 gcc_assert (entry); | |
| 1289 | |
| 1290 /* Calculate the index of the object on the page; this is its bit | |
| 1291 position in the in_use_p bitmap. Note that because a char* might | |
| 1292 point to the middle of an object, we need special code here to | |
| 1293 make sure P points to the start of an object. */ | |
| 1294 offset = ((const char *) p - entry->page) % object_size_table[entry->order]; | |
| 1295 if (offset) | |
| 1296 { | |
| 1297 /* Here we've seen a char* which does not point to the beginning | |
| 1298 of an allocated object. We assume it points to the middle of | |
| 1299 a STRING_CST. */ | |
| 1300 gcc_assert (offset == offsetof (struct tree_string, str)); | |
| 1301 p = ((const char *) p) - offset; | |
| 1302 gt_ggc_mx_lang_tree_node (CONST_CAST (void *, p)); | |
| 1303 return; | |
| 1304 } | |
| 1305 | |
| 1306 bit = OFFSET_TO_BIT (((const char *) p) - entry->page, entry->order); | |
| 1307 word = bit / HOST_BITS_PER_LONG; | |
| 1308 mask = (unsigned long) 1 << (bit % HOST_BITS_PER_LONG); | |
| 1309 | |
| 1310 /* If the bit was previously set, skip it. */ | |
| 1311 if (entry->in_use_p[word] & mask) | |
| 1312 return; | |
| 1313 | |
| 1314 /* Otherwise set it, and decrement the free object count. */ | |
| 1315 entry->in_use_p[word] |= mask; | |
| 1316 entry->num_free_objects -= 1; | |
| 1317 | |
| 1318 if (GGC_DEBUG_LEVEL >= 4) | |
| 1319 fprintf (G.debug_file, "Marking %p\n", p); | |
| 1320 | |
| 1321 return; | |
| 1322 } | |
| 1323 | |
| 1324 /* If P is not marked, marks it and return false. Otherwise return true. | |
| 1325 P must have been allocated by the GC allocator; it mustn't point to | |
| 1326 static objects, stack variables, or memory allocated with malloc. */ | |
| 1327 | |
| 1328 int | |
| 1329 ggc_set_mark (const void *p) | |
| 1330 { | |
| 1331 page_entry *entry; | |
| 1332 unsigned bit, word; | |
| 1333 unsigned long mask; | |
| 1334 | |
| 1335 /* Look up the page on which the object is alloced. If the object | |
| 1336 wasn't allocated by the collector, we'll probably die. */ | |
| 1337 entry = lookup_page_table_entry (p); | |
| 1338 gcc_assert (entry); | |
| 1339 | |
| 1340 /* Calculate the index of the object on the page; this is its bit | |
| 1341 position in the in_use_p bitmap. */ | |
| 1342 bit = OFFSET_TO_BIT (((const char *) p) - entry->page, entry->order); | |
| 1343 word = bit / HOST_BITS_PER_LONG; | |
| 1344 mask = (unsigned long) 1 << (bit % HOST_BITS_PER_LONG); | |
| 1345 | |
| 1346 /* If the bit was previously set, skip it. */ | |
| 1347 if (entry->in_use_p[word] & mask) | |
| 1348 return 1; | |
| 1349 | |
| 1350 /* Otherwise set it, and decrement the free object count. */ | |
| 1351 entry->in_use_p[word] |= mask; | |
| 1352 entry->num_free_objects -= 1; | |
| 1353 | |
| 1354 if (GGC_DEBUG_LEVEL >= 4) | |
| 1355 fprintf (G.debug_file, "Marking %p\n", p); | |
| 1356 | |
| 1357 return 0; | |
| 1358 } | |
| 1359 | |
| 1360 /* Return 1 if P has been marked, zero otherwise. | |
| 1361 P must have been allocated by the GC allocator; it mustn't point to | |
| 1362 static objects, stack variables, or memory allocated with malloc. */ | |
| 1363 | |
| 1364 int | |
| 1365 ggc_marked_p (const void *p) | |
| 1366 { | |
| 1367 page_entry *entry; | |
| 1368 unsigned bit, word; | |
| 1369 unsigned long mask; | |
| 1370 | |
| 1371 /* Look up the page on which the object is alloced. If the object | |
| 1372 wasn't allocated by the collector, we'll probably die. */ | |
| 1373 entry = lookup_page_table_entry (p); | |
| 1374 gcc_assert (entry); | |
| 1375 | |
| 1376 /* Calculate the index of the object on the page; this is its bit | |
| 1377 position in the in_use_p bitmap. */ | |
| 1378 bit = OFFSET_TO_BIT (((const char *) p) - entry->page, entry->order); | |
| 1379 word = bit / HOST_BITS_PER_LONG; | |
| 1380 mask = (unsigned long) 1 << (bit % HOST_BITS_PER_LONG); | |
| 1381 | |
| 1382 return (entry->in_use_p[word] & mask) != 0; | |
| 1383 } | |
| 1384 | |
| 1385 /* Return the size of the gc-able object P. */ | |
| 1386 | |
| 1387 size_t | |
| 1388 ggc_get_size (const void *p) | |
| 1389 { | |
| 1390 page_entry *pe = lookup_page_table_entry (p); | |
| 1391 return OBJECT_SIZE (pe->order); | |
| 1392 } | |
| 1393 | |
| 1394 /* Release the memory for object P. */ | |
| 1395 | |
| 1396 void | |
| 1397 ggc_free (void *p) | |
| 1398 { | |
| 1399 page_entry *pe = lookup_page_table_entry (p); | |
| 1400 size_t order = pe->order; | |
| 1401 size_t size = OBJECT_SIZE (order); | |
| 1402 | |
| 1403 #ifdef GATHER_STATISTICS | |
| 1404 ggc_free_overhead (p); | |
| 1405 #endif | |
| 1406 | |
| 1407 if (GGC_DEBUG_LEVEL >= 3) | |
| 1408 fprintf (G.debug_file, | |
| 1409 "Freeing object, actual size=%lu, at %p on %p\n", | |
| 1410 (unsigned long) size, p, (void *) pe); | |
| 1411 | |
| 1412 #ifdef ENABLE_GC_CHECKING | |
| 1413 /* Poison the data, to indicate the data is garbage. */ | |
| 1414 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (p, size)); | |
| 1415 memset (p, 0xa5, size); | |
| 1416 #endif | |
| 1417 /* Let valgrind know the object is free. */ | |
| 1418 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (p, size)); | |
| 1419 | |
| 1420 #ifdef ENABLE_GC_ALWAYS_COLLECT | |
| 1421 /* 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
|
1422 the data, but instead verify that the data is *actually* not |
| 0 | 1423 reachable the next time we collect. */ |
| 1424 { | |
| 1425 struct free_object *fo = XNEW (struct free_object); | |
| 1426 fo->object = p; | |
| 1427 fo->next = G.free_object_list; | |
| 1428 G.free_object_list = fo; | |
| 1429 } | |
| 1430 #else | |
| 1431 { | |
| 1432 unsigned int bit_offset, word, bit; | |
| 1433 | |
| 1434 G.allocated -= size; | |
| 1435 | |
| 1436 /* Mark the object not-in-use. */ | |
| 1437 bit_offset = OFFSET_TO_BIT (((const char *) p) - pe->page, order); | |
| 1438 word = bit_offset / HOST_BITS_PER_LONG; | |
| 1439 bit = bit_offset % HOST_BITS_PER_LONG; | |
| 1440 pe->in_use_p[word] &= ~(1UL << bit); | |
| 1441 | |
| 1442 if (pe->num_free_objects++ == 0) | |
| 1443 { | |
| 1444 page_entry *p, *q; | |
| 1445 | |
| 1446 /* If the page is completely full, then it's supposed to | |
| 1447 be after all pages that aren't. Since we've freed one | |
| 1448 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
|
1449 page to the head of the list. |
| 0 | 1450 |
| 1451 PE is the node we want to move. Q is the previous node | |
| 1452 and P is the next node in the list. */ | |
| 1453 q = pe->prev; | |
| 1454 if (q && q->num_free_objects == 0) | |
| 1455 { | |
| 1456 p = pe->next; | |
| 1457 | |
| 1458 q->next = p; | |
| 1459 | |
| 1460 /* If PE was at the end of the list, then Q becomes the | |
| 1461 new end of the list. If PE was not the end of the | |
| 1462 list, then we need to update the PREV field for P. */ | |
| 1463 if (!p) | |
| 1464 G.page_tails[order] = q; | |
| 1465 else | |
| 1466 p->prev = q; | |
| 1467 | |
| 1468 /* Move PE to the head of the list. */ | |
| 1469 pe->next = G.pages[order]; | |
| 1470 pe->prev = NULL; | |
| 1471 G.pages[order]->prev = pe; | |
| 1472 G.pages[order] = pe; | |
| 1473 } | |
| 1474 | |
| 1475 /* Reset the hint bit to point to the only free object. */ | |
| 1476 pe->next_bit_hint = bit_offset; | |
| 1477 } | |
| 1478 } | |
| 1479 #endif | |
| 1480 } | |
| 1481 | |
| 1482 /* Subroutine of init_ggc which computes the pair of numbers used to | |
| 1483 perform division by OBJECT_SIZE (order) and fills in inverse_table[]. | |
| 1484 | |
| 1485 This algorithm is taken from Granlund and Montgomery's paper | |
| 1486 "Division by Invariant Integers using Multiplication" | |
| 1487 (Proc. SIGPLAN PLDI, 1994), section 9 (Exact division by | |
| 1488 constants). */ | |
| 1489 | |
| 1490 static void | |
| 1491 compute_inverse (unsigned order) | |
| 1492 { | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1493 size_t size, inv; |
| 0 | 1494 unsigned int e; |
| 1495 | |
| 1496 size = OBJECT_SIZE (order); | |
| 1497 e = 0; | |
| 1498 while (size % 2 == 0) | |
| 1499 { | |
| 1500 e++; | |
| 1501 size >>= 1; | |
| 1502 } | |
| 1503 | |
| 1504 inv = size; | |
| 1505 while (inv * size != 1) | |
| 1506 inv = inv * (2 - inv*size); | |
| 1507 | |
| 1508 DIV_MULT (order) = inv; | |
| 1509 DIV_SHIFT (order) = e; | |
| 1510 } | |
| 1511 | |
| 1512 /* Initialize the ggc-mmap allocator. */ | |
| 1513 void | |
| 1514 init_ggc (void) | |
| 1515 { | |
| 1516 unsigned order; | |
| 1517 | |
| 1518 G.pagesize = getpagesize(); | |
| 1519 G.lg_pagesize = exact_log2 (G.pagesize); | |
| 1520 | |
| 1521 #ifdef HAVE_MMAP_DEV_ZERO | |
| 1522 G.dev_zero_fd = open ("/dev/zero", O_RDONLY); | |
| 1523 if (G.dev_zero_fd == -1) | |
| 1524 internal_error ("open /dev/zero: %m"); | |
| 1525 #endif | |
| 1526 | |
| 1527 #if 0 | |
| 1528 G.debug_file = fopen ("ggc-mmap.debug", "w"); | |
| 1529 #else | |
| 1530 G.debug_file = stdout; | |
| 1531 #endif | |
| 1532 | |
| 1533 #ifdef USING_MMAP | |
| 1534 /* StunOS has an amazing off-by-one error for the first mmap allocation | |
| 1535 after fiddling with RLIMIT_STACK. The result, as hard as it is to | |
| 1536 believe, is an unaligned page allocation, which would cause us to | |
| 1537 hork badly if we tried to use it. */ | |
| 1538 { | |
| 1539 char *p = alloc_anon (NULL, G.pagesize); | |
| 1540 struct page_entry *e; | |
| 1541 if ((size_t)p & (G.pagesize - 1)) | |
| 1542 { | |
| 1543 /* How losing. Discard this one and try another. If we still | |
| 1544 can't get something useful, give up. */ | |
| 1545 | |
| 1546 p = alloc_anon (NULL, G.pagesize); | |
| 1547 gcc_assert (!((size_t)p & (G.pagesize - 1))); | |
| 1548 } | |
| 1549 | |
| 1550 /* We have a good page, might as well hold onto it... */ | |
| 1551 e = XCNEW (struct page_entry); | |
| 1552 e->bytes = G.pagesize; | |
| 1553 e->page = p; | |
| 1554 e->next = G.free_pages; | |
| 1555 G.free_pages = e; | |
| 1556 } | |
| 1557 #endif | |
| 1558 | |
| 1559 /* Initialize the object size table. */ | |
| 1560 for (order = 0; order < HOST_BITS_PER_PTR; ++order) | |
| 1561 object_size_table[order] = (size_t) 1 << order; | |
| 1562 for (order = HOST_BITS_PER_PTR; order < NUM_ORDERS; ++order) | |
| 1563 { | |
| 1564 size_t s = extra_order_size_table[order - HOST_BITS_PER_PTR]; | |
| 1565 | |
| 1566 /* If S is not a multiple of the MAX_ALIGNMENT, then round it up | |
| 1567 so that we're sure of getting aligned memory. */ | |
| 1568 s = ROUND_UP (s, MAX_ALIGNMENT); | |
| 1569 object_size_table[order] = s; | |
| 1570 } | |
| 1571 | |
| 1572 /* Initialize the objects-per-page and inverse tables. */ | |
| 1573 for (order = 0; order < NUM_ORDERS; ++order) | |
| 1574 { | |
| 1575 objects_per_page_table[order] = G.pagesize / OBJECT_SIZE (order); | |
| 1576 if (objects_per_page_table[order] == 0) | |
| 1577 objects_per_page_table[order] = 1; | |
| 1578 compute_inverse (order); | |
| 1579 } | |
| 1580 | |
| 1581 /* Reset the size_lookup array to put appropriately sized objects in | |
| 1582 the special orders. All objects bigger than the previous power | |
| 1583 of two, but no greater than the special size, should go in the | |
| 1584 new order. */ | |
| 1585 for (order = HOST_BITS_PER_PTR; order < NUM_ORDERS; ++order) | |
| 1586 { | |
| 1587 int o; | |
| 1588 int i; | |
| 1589 | |
| 1590 i = OBJECT_SIZE (order); | |
| 1591 if (i >= NUM_SIZE_LOOKUP) | |
| 1592 continue; | |
| 1593 | |
| 1594 for (o = size_lookup[i]; o == size_lookup [i]; --i) | |
| 1595 size_lookup[i] = order; | |
| 1596 } | |
| 1597 | |
| 1598 G.depth_in_use = 0; | |
| 1599 G.depth_max = 10; | |
| 1600 G.depth = XNEWVEC (unsigned int, G.depth_max); | |
| 1601 | |
| 1602 G.by_depth_in_use = 0; | |
| 1603 G.by_depth_max = INITIAL_PTE_COUNT; | |
| 1604 G.by_depth = XNEWVEC (page_entry *, G.by_depth_max); | |
| 1605 G.save_in_use = XNEWVEC (unsigned long *, G.by_depth_max); | |
| 1606 } | |
| 1607 | |
| 1608 /* Start a new GGC zone. */ | |
| 1609 | |
| 1610 struct alloc_zone * | |
| 1611 new_ggc_zone (const char *name ATTRIBUTE_UNUSED) | |
| 1612 { | |
| 1613 return NULL; | |
| 1614 } | |
| 1615 | |
| 1616 /* Destroy a GGC zone. */ | |
| 1617 void | |
| 1618 destroy_ggc_zone (struct alloc_zone *zone ATTRIBUTE_UNUSED) | |
| 1619 { | |
| 1620 } | |
| 1621 | |
| 1622 /* Merge the SAVE_IN_USE_P and IN_USE_P arrays in P so that IN_USE_P | |
| 1623 reflects reality. Recalculate NUM_FREE_OBJECTS as well. */ | |
| 1624 | |
| 1625 static void | |
| 1626 ggc_recalculate_in_use_p (page_entry *p) | |
| 1627 { | |
| 1628 unsigned int i; | |
| 1629 size_t num_objects; | |
| 1630 | |
| 1631 /* Because the past-the-end bit in in_use_p is always set, we | |
| 1632 pretend there is one additional object. */ | |
| 1633 num_objects = OBJECTS_IN_PAGE (p) + 1; | |
| 1634 | |
| 1635 /* Reset the free object count. */ | |
| 1636 p->num_free_objects = num_objects; | |
| 1637 | |
| 1638 /* Combine the IN_USE_P and SAVE_IN_USE_P arrays. */ | |
| 1639 for (i = 0; | |
| 1640 i < CEIL (BITMAP_SIZE (num_objects), | |
| 1641 sizeof (*p->in_use_p)); | |
| 1642 ++i) | |
| 1643 { | |
| 1644 unsigned long j; | |
| 1645 | |
| 1646 /* Something is in use if it is marked, or if it was in use in a | |
| 1647 context further down the context stack. */ | |
| 1648 p->in_use_p[i] |= save_in_use_p (p)[i]; | |
| 1649 | |
| 1650 /* Decrement the free object count for every object allocated. */ | |
| 1651 for (j = p->in_use_p[i]; j; j >>= 1) | |
| 1652 p->num_free_objects -= (j & 1); | |
| 1653 } | |
| 1654 | |
| 1655 gcc_assert (p->num_free_objects < num_objects); | |
| 1656 } | |
| 1657 | |
| 1658 /* Unmark all objects. */ | |
| 1659 | |
| 1660 static void | |
| 1661 clear_marks (void) | |
| 1662 { | |
| 1663 unsigned order; | |
| 1664 | |
| 1665 for (order = 2; order < NUM_ORDERS; order++) | |
| 1666 { | |
| 1667 page_entry *p; | |
| 1668 | |
| 1669 for (p = G.pages[order]; p != NULL; p = p->next) | |
| 1670 { | |
| 1671 size_t num_objects = OBJECTS_IN_PAGE (p); | |
| 1672 size_t bitmap_size = BITMAP_SIZE (num_objects + 1); | |
| 1673 | |
| 1674 /* The data should be page-aligned. */ | |
| 1675 gcc_assert (!((size_t) p->page & (G.pagesize - 1))); | |
| 1676 | |
| 1677 /* Pages that aren't in the topmost context are not collected; | |
| 1678 nevertheless, we need their in-use bit vectors to store GC | |
| 1679 marks. So, back them up first. */ | |
| 1680 if (p->context_depth < G.context_depth) | |
| 1681 { | |
| 1682 if (! save_in_use_p (p)) | |
| 1683 save_in_use_p (p) = XNEWVAR (unsigned long, bitmap_size); | |
| 1684 memcpy (save_in_use_p (p), p->in_use_p, bitmap_size); | |
| 1685 } | |
| 1686 | |
| 1687 /* Reset reset the number of free objects and clear the | |
| 1688 in-use bits. These will be adjusted by mark_obj. */ | |
| 1689 p->num_free_objects = num_objects; | |
| 1690 memset (p->in_use_p, 0, bitmap_size); | |
| 1691 | |
| 1692 /* Make sure the one-past-the-end bit is always set. */ | |
| 1693 p->in_use_p[num_objects / HOST_BITS_PER_LONG] | |
| 1694 = ((unsigned long) 1 << (num_objects % HOST_BITS_PER_LONG)); | |
| 1695 } | |
| 1696 } | |
| 1697 } | |
| 1698 | |
| 1699 /* Free all empty pages. Partially empty pages need no attention | |
| 1700 because the `mark' bit doubles as an `unused' bit. */ | |
| 1701 | |
| 1702 static void | |
| 1703 sweep_pages (void) | |
| 1704 { | |
| 1705 unsigned order; | |
| 1706 | |
| 1707 for (order = 2; order < NUM_ORDERS; order++) | |
| 1708 { | |
| 1709 /* The last page-entry to consider, regardless of entries | |
| 1710 placed at the end of the list. */ | |
| 1711 page_entry * const last = G.page_tails[order]; | |
| 1712 | |
| 1713 size_t num_objects; | |
| 1714 size_t live_objects; | |
| 1715 page_entry *p, *previous; | |
| 1716 int done; | |
| 1717 | |
| 1718 p = G.pages[order]; | |
| 1719 if (p == NULL) | |
| 1720 continue; | |
| 1721 | |
| 1722 previous = NULL; | |
| 1723 do | |
| 1724 { | |
| 1725 page_entry *next = p->next; | |
| 1726 | |
| 1727 /* Loop until all entries have been examined. */ | |
| 1728 done = (p == last); | |
| 1729 | |
| 1730 num_objects = OBJECTS_IN_PAGE (p); | |
| 1731 | |
| 1732 /* Add all live objects on this page to the count of | |
| 1733 allocated memory. */ | |
| 1734 live_objects = num_objects - p->num_free_objects; | |
| 1735 | |
| 1736 G.allocated += OBJECT_SIZE (order) * live_objects; | |
| 1737 | |
| 1738 /* Only objects on pages in the topmost context should get | |
| 1739 collected. */ | |
| 1740 if (p->context_depth < G.context_depth) | |
| 1741 ; | |
| 1742 | |
| 1743 /* Remove the page if it's empty. */ | |
| 1744 else if (live_objects == 0) | |
| 1745 { | |
| 1746 /* If P was the first page in the list, then NEXT | |
| 1747 becomes the new first page in the list, otherwise | |
| 1748 splice P out of the forward pointers. */ | |
| 1749 if (! previous) | |
| 1750 G.pages[order] = next; | |
| 1751 else | |
| 1752 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
|
1753 |
| 0 | 1754 /* Splice P out of the back pointers too. */ |
| 1755 if (next) | |
| 1756 next->prev = previous; | |
| 1757 | |
| 1758 /* Are we removing the last element? */ | |
| 1759 if (p == G.page_tails[order]) | |
| 1760 G.page_tails[order] = previous; | |
| 1761 free_page (p); | |
| 1762 p = previous; | |
| 1763 } | |
| 1764 | |
| 1765 /* If the page is full, move it to the end. */ | |
| 1766 else if (p->num_free_objects == 0) | |
| 1767 { | |
| 1768 /* Don't move it if it's already at the end. */ | |
| 1769 if (p != G.page_tails[order]) | |
| 1770 { | |
| 1771 /* Move p to the end of the list. */ | |
| 1772 p->next = NULL; | |
| 1773 p->prev = G.page_tails[order]; | |
| 1774 G.page_tails[order]->next = p; | |
| 1775 | |
| 1776 /* Update the tail pointer... */ | |
| 1777 G.page_tails[order] = p; | |
| 1778 | |
| 1779 /* ... and the head pointer, if necessary. */ | |
| 1780 if (! previous) | |
| 1781 G.pages[order] = next; | |
| 1782 else | |
| 1783 previous->next = next; | |
| 1784 | |
| 1785 /* And update the backpointer in NEXT if necessary. */ | |
| 1786 if (next) | |
| 1787 next->prev = previous; | |
| 1788 | |
| 1789 p = previous; | |
| 1790 } | |
| 1791 } | |
| 1792 | |
| 1793 /* If we've fallen through to here, it's a page in the | |
| 1794 topmost context that is neither full nor empty. Such a | |
| 1795 page must precede pages at lesser context depth in the | |
| 1796 list, so move it to the head. */ | |
| 1797 else if (p != G.pages[order]) | |
| 1798 { | |
| 1799 previous->next = p->next; | |
| 1800 | |
| 1801 /* Update the backchain in the next node if it exists. */ | |
| 1802 if (p->next) | |
| 1803 p->next->prev = previous; | |
| 1804 | |
| 1805 /* Move P to the head of the list. */ | |
| 1806 p->next = G.pages[order]; | |
| 1807 p->prev = NULL; | |
| 1808 G.pages[order]->prev = p; | |
| 1809 | |
| 1810 /* Update the head pointer. */ | |
| 1811 G.pages[order] = p; | |
| 1812 | |
| 1813 /* Are we moving the last element? */ | |
| 1814 if (G.page_tails[order] == p) | |
| 1815 G.page_tails[order] = previous; | |
| 1816 p = previous; | |
| 1817 } | |
| 1818 | |
| 1819 previous = p; | |
| 1820 p = next; | |
| 1821 } | |
| 1822 while (! done); | |
| 1823 | |
| 1824 /* Now, restore the in_use_p vectors for any pages from contexts | |
| 1825 other than the current one. */ | |
| 1826 for (p = G.pages[order]; p; p = p->next) | |
| 1827 if (p->context_depth != G.context_depth) | |
| 1828 ggc_recalculate_in_use_p (p); | |
| 1829 } | |
| 1830 } | |
| 1831 | |
| 1832 #ifdef ENABLE_GC_CHECKING | |
| 1833 /* Clobber all free objects. */ | |
| 1834 | |
| 1835 static void | |
| 1836 poison_pages (void) | |
| 1837 { | |
| 1838 unsigned order; | |
| 1839 | |
| 1840 for (order = 2; order < NUM_ORDERS; order++) | |
| 1841 { | |
| 1842 size_t size = OBJECT_SIZE (order); | |
| 1843 page_entry *p; | |
| 1844 | |
| 1845 for (p = G.pages[order]; p != NULL; p = p->next) | |
| 1846 { | |
| 1847 size_t num_objects; | |
| 1848 size_t i; | |
| 1849 | |
| 1850 if (p->context_depth != G.context_depth) | |
| 1851 /* Since we don't do any collection for pages in pushed | |
| 1852 contexts, there's no need to do any poisoning. And | |
| 1853 besides, the IN_USE_P array isn't valid until we pop | |
| 1854 contexts. */ | |
| 1855 continue; | |
| 1856 | |
| 1857 num_objects = OBJECTS_IN_PAGE (p); | |
| 1858 for (i = 0; i < num_objects; i++) | |
| 1859 { | |
| 1860 size_t word, bit; | |
| 1861 word = i / HOST_BITS_PER_LONG; | |
| 1862 bit = i % HOST_BITS_PER_LONG; | |
| 1863 if (((p->in_use_p[word] >> bit) & 1) == 0) | |
| 1864 { | |
| 1865 char *object = p->page + i * size; | |
| 1866 | |
| 1867 /* Keep poison-by-write when we expect to use Valgrind, | |
| 1868 so the exact same memory semantics is kept, in case | |
| 1869 there are memory errors. We override this request | |
| 1870 below. */ | |
| 1871 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_UNDEFINED (object, | |
| 1872 size)); | |
| 1873 memset (object, 0xa5, size); | |
| 1874 | |
| 1875 /* Drop the handle to avoid handle leak. */ | |
| 1876 VALGRIND_DISCARD (VALGRIND_MAKE_MEM_NOACCESS (object, size)); | |
| 1877 } | |
| 1878 } | |
| 1879 } | |
| 1880 } | |
| 1881 } | |
| 1882 #else | |
| 1883 #define poison_pages() | |
| 1884 #endif | |
| 1885 | |
| 1886 #ifdef ENABLE_GC_ALWAYS_COLLECT | |
| 1887 /* Validate that the reportedly free objects actually are. */ | |
| 1888 | |
| 1889 static void | |
| 1890 validate_free_objects (void) | |
| 1891 { | |
| 1892 struct free_object *f, *next, *still_free = NULL; | |
| 1893 | |
| 1894 for (f = G.free_object_list; f ; f = next) | |
| 1895 { | |
| 1896 page_entry *pe = lookup_page_table_entry (f->object); | |
| 1897 size_t bit, word; | |
| 1898 | |
| 1899 bit = OFFSET_TO_BIT ((char *)f->object - pe->page, pe->order); | |
| 1900 word = bit / HOST_BITS_PER_LONG; | |
| 1901 bit = bit % HOST_BITS_PER_LONG; | |
| 1902 next = f->next; | |
| 1903 | |
| 1904 /* Make certain it isn't visible from any root. Notice that we | |
| 1905 do this check before sweep_pages merges save_in_use_p. */ | |
| 1906 gcc_assert (!(pe->in_use_p[word] & (1UL << bit))); | |
| 1907 | |
| 1908 /* If the object comes from an outer context, then retain the | |
| 1909 free_object entry, so that we can verify that the address | |
| 1910 isn't live on the stack in some outer context. */ | |
| 1911 if (pe->context_depth != G.context_depth) | |
| 1912 { | |
| 1913 f->next = still_free; | |
| 1914 still_free = f; | |
| 1915 } | |
| 1916 else | |
| 1917 free (f); | |
| 1918 } | |
| 1919 | |
| 1920 G.free_object_list = still_free; | |
| 1921 } | |
| 1922 #else | |
| 1923 #define validate_free_objects() | |
| 1924 #endif | |
| 1925 | |
| 1926 /* Top level mark-and-sweep routine. */ | |
| 1927 | |
| 1928 void | |
| 1929 ggc_collect (void) | |
| 1930 { | |
| 1931 /* Avoid frequent unnecessary work by skipping collection if the | |
| 1932 total allocations haven't expanded much since the last | |
| 1933 collection. */ | |
| 1934 float allocated_last_gc = | |
| 1935 MAX (G.allocated_last_gc, (size_t)PARAM_VALUE (GGC_MIN_HEAPSIZE) * 1024); | |
| 1936 | |
| 1937 float min_expand = allocated_last_gc * PARAM_VALUE (GGC_MIN_EXPAND) / 100; | |
| 1938 | |
| 1939 if (G.allocated < allocated_last_gc + min_expand && !ggc_force_collect) | |
| 1940 return; | |
| 1941 | |
| 1942 timevar_push (TV_GC); | |
| 1943 if (!quiet_flag) | |
| 1944 fprintf (stderr, " {GC %luk -> ", (unsigned long) G.allocated / 1024); | |
| 1945 if (GGC_DEBUG_LEVEL >= 2) | |
| 1946 fprintf (G.debug_file, "BEGIN COLLECTING\n"); | |
| 1947 | |
| 1948 /* Zero the total allocated bytes. This will be recalculated in the | |
| 1949 sweep phase. */ | |
| 1950 G.allocated = 0; | |
| 1951 | |
| 1952 /* Release the pages we freed the last time we collected, but didn't | |
| 1953 reuse in the interim. */ | |
| 1954 release_pages (); | |
| 1955 | |
| 1956 /* Indicate that we've seen collections at this context depth. */ | |
| 1957 G.context_depth_collections = ((unsigned long)1 << (G.context_depth + 1)) - 1; | |
| 1958 | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1959 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
|
1960 |
| 0 | 1961 clear_marks (); |
| 1962 ggc_mark_roots (); | |
| 1963 #ifdef GATHER_STATISTICS | |
| 1964 ggc_prune_overhead_list (); | |
| 1965 #endif | |
| 1966 poison_pages (); | |
| 1967 validate_free_objects (); | |
| 1968 sweep_pages (); | |
| 1969 | |
| 1970 G.allocated_last_gc = G.allocated; | |
| 1971 | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
1972 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
|
1973 |
| 0 | 1974 timevar_pop (TV_GC); |
| 1975 | |
| 1976 if (!quiet_flag) | |
| 1977 fprintf (stderr, "%luk}", (unsigned long) G.allocated / 1024); | |
| 1978 if (GGC_DEBUG_LEVEL >= 2) | |
| 1979 fprintf (G.debug_file, "END COLLECTING\n"); | |
| 1980 } | |
| 1981 | |
| 1982 /* Print allocation statistics. */ | |
| 1983 #define SCALE(x) ((unsigned long) ((x) < 1024*10 \ | |
| 1984 ? (x) \ | |
| 1985 : ((x) < 1024*1024*10 \ | |
| 1986 ? (x) / 1024 \ | |
| 1987 : (x) / (1024*1024)))) | |
| 1988 #define STAT_LABEL(x) ((x) < 1024*10 ? ' ' : ((x) < 1024*1024*10 ? 'k' : 'M')) | |
| 1989 | |
| 1990 void | |
| 1991 ggc_print_statistics (void) | |
| 1992 { | |
| 1993 struct ggc_statistics stats; | |
| 1994 unsigned int i; | |
| 1995 size_t total_overhead = 0; | |
| 1996 | |
| 1997 /* Clear the statistics. */ | |
| 1998 memset (&stats, 0, sizeof (stats)); | |
| 1999 | |
| 2000 /* Make sure collection will really occur. */ | |
| 2001 G.allocated_last_gc = 0; | |
| 2002 | |
| 2003 /* Collect and print the statistics common across collectors. */ | |
| 2004 ggc_print_common_statistics (stderr, &stats); | |
| 2005 | |
| 2006 /* Release free pages so that we will not count the bytes allocated | |
| 2007 there as part of the total allocated memory. */ | |
| 2008 release_pages (); | |
| 2009 | |
| 2010 /* Collect some information about the various sizes of | |
| 2011 allocation. */ | |
| 2012 fprintf (stderr, | |
| 2013 "Memory still allocated at the end of the compilation process\n"); | |
| 2014 fprintf (stderr, "%-5s %10s %10s %10s\n", | |
| 2015 "Size", "Allocated", "Used", "Overhead"); | |
| 2016 for (i = 0; i < NUM_ORDERS; ++i) | |
| 2017 { | |
| 2018 page_entry *p; | |
| 2019 size_t allocated; | |
| 2020 size_t in_use; | |
| 2021 size_t overhead; | |
| 2022 | |
| 2023 /* Skip empty entries. */ | |
| 2024 if (!G.pages[i]) | |
| 2025 continue; | |
| 2026 | |
| 2027 overhead = allocated = in_use = 0; | |
| 2028 | |
| 2029 /* Figure out the total number of bytes allocated for objects of | |
| 2030 this size, and how many of them are actually in use. Also figure | |
| 2031 out how much memory the page table is using. */ | |
| 2032 for (p = G.pages[i]; p; p = p->next) | |
| 2033 { | |
| 2034 allocated += p->bytes; | |
| 2035 in_use += | |
| 2036 (OBJECTS_IN_PAGE (p) - p->num_free_objects) * OBJECT_SIZE (i); | |
| 2037 | |
| 2038 overhead += (sizeof (page_entry) - sizeof (long) | |
| 2039 + BITMAP_SIZE (OBJECTS_IN_PAGE (p) + 1)); | |
| 2040 } | |
| 2041 fprintf (stderr, "%-5lu %10lu%c %10lu%c %10lu%c\n", | |
| 2042 (unsigned long) OBJECT_SIZE (i), | |
| 2043 SCALE (allocated), STAT_LABEL (allocated), | |
| 2044 SCALE (in_use), STAT_LABEL (in_use), | |
| 2045 SCALE (overhead), STAT_LABEL (overhead)); | |
| 2046 total_overhead += overhead; | |
| 2047 } | |
| 2048 fprintf (stderr, "%-5s %10lu%c %10lu%c %10lu%c\n", "Total", | |
| 2049 SCALE (G.bytes_mapped), STAT_LABEL (G.bytes_mapped), | |
| 2050 SCALE (G.allocated), STAT_LABEL(G.allocated), | |
| 2051 SCALE (total_overhead), STAT_LABEL (total_overhead)); | |
| 2052 | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2053 #ifdef GATHER_STATISTICS |
| 0 | 2054 { |
| 2055 fprintf (stderr, "\nTotal allocations and overheads during the compilation process\n"); | |
| 2056 | |
| 2057 fprintf (stderr, "Total Overhead: %10lld\n", | |
| 2058 G.stats.total_overhead); | |
| 2059 fprintf (stderr, "Total Allocated: %10lld\n", | |
| 2060 G.stats.total_allocated); | |
| 2061 | |
| 2062 fprintf (stderr, "Total Overhead under 32B: %10lld\n", | |
| 2063 G.stats.total_overhead_under32); | |
| 2064 fprintf (stderr, "Total Allocated under 32B: %10lld\n", | |
| 2065 G.stats.total_allocated_under32); | |
| 2066 fprintf (stderr, "Total Overhead under 64B: %10lld\n", | |
| 2067 G.stats.total_overhead_under64); | |
| 2068 fprintf (stderr, "Total Allocated under 64B: %10lld\n", | |
| 2069 G.stats.total_allocated_under64); | |
| 2070 fprintf (stderr, "Total Overhead under 128B: %10lld\n", | |
| 2071 G.stats.total_overhead_under128); | |
| 2072 fprintf (stderr, "Total Allocated under 128B: %10lld\n", | |
| 2073 G.stats.total_allocated_under128); | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2074 |
| 0 | 2075 for (i = 0; i < NUM_ORDERS; i++) |
| 2076 if (G.stats.total_allocated_per_order[i]) | |
| 2077 { | |
| 2078 fprintf (stderr, "Total Overhead page size %7lu: %10lld\n", | |
| 2079 (unsigned long) OBJECT_SIZE (i), | |
| 2080 G.stats.total_overhead_per_order[i]); | |
| 2081 fprintf (stderr, "Total Allocated page size %7lu: %10lld\n", | |
| 2082 (unsigned long) OBJECT_SIZE (i), | |
| 2083 G.stats.total_allocated_per_order[i]); | |
| 2084 } | |
| 2085 } | |
| 2086 #endif | |
| 2087 } | |
| 2088 | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2089 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
|
2090 { |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2091 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
|
2092 }; |
|
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2093 |
| 0 | 2094 struct ggc_pch_data |
| 2095 { | |
|
55
77e2b8dfacca
update it from 4.4.3 to 4.5.0
ryoma <e075725@ie.u-ryukyu.ac.jp>
parents:
0
diff
changeset
|
2096 struct ggc_pch_ondisk d; |
| 0 | 2097 size_t base[NUM_ORDERS]; |
| 2098 size_t written[NUM_ORDERS]; | |
| 2099 }; | |
| 2100 | |
| 2101 struct ggc_pch_data * | |
| 2102 init_ggc_pch (void) | |
| 2103 { | |
| 2104 return XCNEW (struct ggc_pch_data); | |
| 2105 } | |
| 2106 | |
| 2107 void | |
| 2108 ggc_pch_count_object (struct ggc_pch_data *d, void *x ATTRIBUTE_UNUSED, | |
| 2109 size_t size, bool is_string ATTRIBUTE_UNUSED, | |
| 2110 enum gt_types_enum type ATTRIBUTE_UNUSED) | |
| 2111 { | |
| 2112 unsigned order; | |
| 2113 | |
| 2114 if (size < NUM_SIZE_LOOKUP) | |
| 2115 order = size_lookup[size]; | |
| 2116 else | |
| 2117 { | |
| 2118 order = 10; | |
| 2119 while (size > OBJECT_SIZE (order)) | |
| 2120 order++; | |
| 2121 } | |
| 2122 | |
| 2123 d->d.totals[order]++; | |
| 2124 } | |
| 2125 | |
| 2126 size_t | |
| 2127 ggc_pch_total_size (struct ggc_pch_data *d) | |
| 2128 { | |
| 2129 size_t a = 0; | |
| 2130 unsigned i; | |
| 2131 | |
| 2132 for (i = 0; i < NUM_ORDERS; i++) | |
| 2133 a += ROUND_UP (d->d.totals[i] * OBJECT_SIZE (i), G.pagesize); | |
| 2134 return a; | |
| 2135 } | |
| 2136 | |
| 2137 void | |
| 2138 ggc_pch_this_base (struct ggc_pch_data *d, void *base) | |
| 2139 { | |
| 2140 size_t a = (size_t) base; | |
| 2141 unsigned i; | |
| 2142 | |
| 2143 for (i = 0; i < NUM_ORDERS; i++) | |
| 2144 { | |
| 2145 d->base[i] = a; | |
| 2146 a += ROUND_UP (d->d.totals[i] * OBJECT_SIZE (i), G.pagesize); | |
| 2147 } | |
| 2148 } | |
| 2149 | |
| 2150 | |
| 2151 char * | |
| 2152 ggc_pch_alloc_object (struct ggc_pch_data *d, void *x ATTRIBUTE_UNUSED, | |
| 2153 size_t size, bool is_string ATTRIBUTE_UNUSED, | |
| 2154 enum gt_types_enum type ATTRIBUTE_UNUSED) | |
| 2155 { | |
| 2156 unsigned order; | |
| 2157 char *result; | |
| 2158 | |
| 2159 if (size < NUM_SIZE_LOOKUP) | |
| 2160 order = size_lookup[size]; | |
| 2161 else | |
| 2162 { | |
| 2163 order = 10; | |
| 2164 while (size > OBJECT_SIZE (order)) | |
| 2165 order++; | |
| 2166 } | |
| 2167 | |
| 2168 result = (char *) d->base[order]; | |
| 2169 d->base[order] += OBJECT_SIZE (order); | |
| 2170 return result; | |
| 2171 } | |
| 2172 | |
| 2173 void | |
| 2174 ggc_pch_prepare_write (struct ggc_pch_data *d ATTRIBUTE_UNUSED, | |
| 2175 FILE *f ATTRIBUTE_UNUSED) | |
| 2176 { | |
| 2177 /* Nothing to do. */ | |
| 2178 } | |
| 2179 | |
| 2180 void | |
| 2181 ggc_pch_write_object (struct ggc_pch_data *d ATTRIBUTE_UNUSED, | |
| 2182 FILE *f, void *x, void *newx ATTRIBUTE_UNUSED, | |
| 2183 size_t size, bool is_string ATTRIBUTE_UNUSED) | |
| 2184 { | |
| 2185 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
|
2186 static const char emptyBytes[256] = { 0 }; |
| 0 | 2187 |
| 2188 if (size < NUM_SIZE_LOOKUP) | |
| 2189 order = size_lookup[size]; | |
| 2190 else | |
| 2191 { | |
| 2192 order = 10; | |
| 2193 while (size > OBJECT_SIZE (order)) | |
| 2194 order++; | |
| 2195 } | |
| 2196 | |
| 2197 if (fwrite (x, size, 1, f) != 1) | |
| 2198 fatal_error ("can't write PCH file: %m"); | |
| 2199 | |
| 2200 /* If SIZE is not the same as OBJECT_SIZE(order), then we need to pad the | |
| 2201 object out to OBJECT_SIZE(order). This happens for strings. */ | |
| 2202 | |
| 2203 if (size != OBJECT_SIZE (order)) | |
| 2204 { | |
| 2205 unsigned padding = OBJECT_SIZE(order) - size; | |
| 2206 | |
| 2207 /* To speed small writes, we use a nulled-out array that's larger | |
| 2208 than most padding requests as the source for our null bytes. This | |
| 2209 permits us to do the padding with fwrite() rather than fseek(), and | |
| 2210 limits the chance the OS may try to flush any outstanding writes. */ | |
| 2211 if (padding <= sizeof(emptyBytes)) | |
| 2212 { | |
| 2213 if (fwrite (emptyBytes, 1, padding, f) != padding) | |
| 2214 fatal_error ("can't write PCH file"); | |
| 2215 } | |
| 2216 else | |
| 2217 { | |
| 2218 /* Larger than our buffer? Just default to fseek. */ | |
| 2219 if (fseek (f, padding, SEEK_CUR) != 0) | |
| 2220 fatal_error ("can't write PCH file"); | |
| 2221 } | |
| 2222 } | |
| 2223 | |
| 2224 d->written[order]++; | |
| 2225 if (d->written[order] == d->d.totals[order] | |
| 2226 && fseek (f, ROUND_UP_VALUE (d->d.totals[order] * OBJECT_SIZE (order), | |
| 2227 G.pagesize), | |
| 2228 SEEK_CUR) != 0) | |
| 2229 fatal_error ("can't write PCH file: %m"); | |
| 2230 } | |
| 2231 | |
| 2232 void | |
| 2233 ggc_pch_finish (struct ggc_pch_data *d, FILE *f) | |
| 2234 { | |
| 2235 if (fwrite (&d->d, sizeof (d->d), 1, f) != 1) | |
| 2236 fatal_error ("can't write PCH file: %m"); | |
| 2237 free (d); | |
| 2238 } | |
| 2239 | |
| 2240 /* Move the PCH PTE entries just added to the end of by_depth, to the | |
| 2241 front. */ | |
| 2242 | |
| 2243 static void | |
| 2244 move_ptes_to_front (int count_old_page_tables, int count_new_page_tables) | |
| 2245 { | |
| 2246 unsigned i; | |
| 2247 | |
| 2248 /* First, we swap the new entries to the front of the varrays. */ | |
| 2249 page_entry **new_by_depth; | |
| 2250 unsigned long **new_save_in_use; | |
| 2251 | |
| 2252 new_by_depth = XNEWVEC (page_entry *, G.by_depth_max); | |
| 2253 new_save_in_use = XNEWVEC (unsigned long *, G.by_depth_max); | |
| 2254 | |
| 2255 memcpy (&new_by_depth[0], | |
| 2256 &G.by_depth[count_old_page_tables], | |
| 2257 count_new_page_tables * sizeof (void *)); | |
| 2258 memcpy (&new_by_depth[count_new_page_tables], | |
| 2259 &G.by_depth[0], | |
| 2260 count_old_page_tables * sizeof (void *)); | |
| 2261 memcpy (&new_save_in_use[0], | |
| 2262 &G.save_in_use[count_old_page_tables], | |
| 2263 count_new_page_tables * sizeof (void *)); | |
| 2264 memcpy (&new_save_in_use[count_new_page_tables], | |
| 2265 &G.save_in_use[0], | |
| 2266 count_old_page_tables * sizeof (void *)); | |
| 2267 | |
| 2268 free (G.by_depth); | |
| 2269 free (G.save_in_use); | |
| 2270 | |
| 2271 G.by_depth = new_by_depth; | |
| 2272 G.save_in_use = new_save_in_use; | |
| 2273 | |
| 2274 /* Now update all the index_by_depth fields. */ | |
| 2275 for (i = G.by_depth_in_use; i > 0; --i) | |
| 2276 { | |
| 2277 page_entry *p = G.by_depth[i-1]; | |
| 2278 p->index_by_depth = i-1; | |
| 2279 } | |
| 2280 | |
| 2281 /* And last, we update the depth pointers in G.depth. The first | |
| 2282 entry is already 0, and context 0 entries always start at index | |
| 2283 0, so there is nothing to update in the first slot. We need a | |
| 2284 second slot, only if we have old ptes, and if we do, they start | |
| 2285 at index count_new_page_tables. */ | |
| 2286 if (count_old_page_tables) | |
| 2287 push_depth (count_new_page_tables); | |
| 2288 } | |
| 2289 | |
| 2290 void | |
| 2291 ggc_pch_read (FILE *f, void *addr) | |
| 2292 { | |
| 2293 struct ggc_pch_ondisk d; | |
| 2294 unsigned i; | |
| 2295 char *offs = (char *) addr; | |
| 2296 unsigned long count_old_page_tables; | |
| 2297 unsigned long count_new_page_tables; | |
| 2298 | |
| 2299 count_old_page_tables = G.by_depth_in_use; | |
| 2300 | |
| 2301 /* We've just read in a PCH file. So, every object that used to be | |
| 2302 allocated is now free. */ | |
| 2303 clear_marks (); | |
| 2304 #ifdef ENABLE_GC_CHECKING | |
| 2305 poison_pages (); | |
| 2306 #endif | |
| 2307 /* Since we free all the allocated objects, the free list becomes | |
| 2308 useless. Validate it now, which will also clear it. */ | |
| 2309 validate_free_objects(); | |
| 2310 | |
| 2311 /* No object read from a PCH file should ever be freed. So, set the | |
| 2312 context depth to 1, and set the depth of all the currently-allocated | |
| 2313 pages to be 1 too. PCH pages will have depth 0. */ | |
| 2314 gcc_assert (!G.context_depth); | |
| 2315 G.context_depth = 1; | |
| 2316 for (i = 0; i < NUM_ORDERS; i++) | |
| 2317 { | |
| 2318 page_entry *p; | |
| 2319 for (p = G.pages[i]; p != NULL; p = p->next) | |
| 2320 p->context_depth = G.context_depth; | |
| 2321 } | |
| 2322 | |
| 2323 /* Allocate the appropriate page-table entries for the pages read from | |
| 2324 the PCH file. */ | |
| 2325 if (fread (&d, sizeof (d), 1, f) != 1) | |
| 2326 fatal_error ("can't read PCH file: %m"); | |
| 2327 | |
| 2328 for (i = 0; i < NUM_ORDERS; i++) | |
| 2329 { | |
| 2330 struct page_entry *entry; | |
| 2331 char *pte; | |
| 2332 size_t bytes; | |
| 2333 size_t num_objs; | |
| 2334 size_t j; | |
| 2335 | |
| 2336 if (d.totals[i] == 0) | |
| 2337 continue; | |
| 2338 | |
| 2339 bytes = ROUND_UP (d.totals[i] * OBJECT_SIZE (i), G.pagesize); | |
| 2340 num_objs = bytes / OBJECT_SIZE (i); | |
| 2341 entry = XCNEWVAR (struct page_entry, (sizeof (struct page_entry) | |
| 2342 - sizeof (long) | |
| 2343 + BITMAP_SIZE (num_objs + 1))); | |
| 2344 entry->bytes = bytes; | |
| 2345 entry->page = offs; | |
| 2346 entry->context_depth = 0; | |
| 2347 offs += bytes; | |
| 2348 entry->num_free_objects = 0; | |
| 2349 entry->order = i; | |
| 2350 | |
| 2351 for (j = 0; | |
| 2352 j + HOST_BITS_PER_LONG <= num_objs + 1; | |
| 2353 j += HOST_BITS_PER_LONG) | |
| 2354 entry->in_use_p[j / HOST_BITS_PER_LONG] = -1; | |
| 2355 for (; j < num_objs + 1; j++) | |
| 2356 entry->in_use_p[j / HOST_BITS_PER_LONG] | |
| 2357 |= 1L << (j % HOST_BITS_PER_LONG); | |
| 2358 | |
| 2359 for (pte = entry->page; | |
| 2360 pte < entry->page + entry->bytes; | |
| 2361 pte += G.pagesize) | |
| 2362 set_page_table_entry (pte, entry); | |
| 2363 | |
| 2364 if (G.page_tails[i] != NULL) | |
| 2365 G.page_tails[i]->next = entry; | |
| 2366 else | |
| 2367 G.pages[i] = entry; | |
| 2368 G.page_tails[i] = entry; | |
| 2369 | |
| 2370 /* We start off by just adding all the new information to the | |
| 2371 end of the varrays, later, we will move the new information | |
| 2372 to the front of the varrays, as the PCH page tables are at | |
| 2373 context 0. */ | |
| 2374 push_by_depth (entry, 0); | |
| 2375 } | |
| 2376 | |
| 2377 /* Now, we update the various data structures that speed page table | |
| 2378 handling. */ | |
| 2379 count_new_page_tables = G.by_depth_in_use - count_old_page_tables; | |
| 2380 | |
| 2381 move_ptes_to_front (count_old_page_tables, count_new_page_tables); | |
| 2382 | |
| 2383 /* Update the statistics. */ | |
| 2384 G.allocated = G.allocated_last_gc = offs - (char *)addr; | |
| 2385 } |