src/share/vm/classfile/symbolTable.hpp

Fri, 29 Aug 2014 13:08:01 +0200

author
mgerdin
date
Fri, 29 Aug 2014 13:08:01 +0200
changeset 7207
152cf4afc11f
parent 6229
5a32d2a3cc1e
child 7535
7ae4e26cb1e0
child 8766
ce9a710b0f63
permissions
-rw-r--r--

8056084: Refactor Hashtable to allow implementations without rehashing support
Reviewed-by: gziemski, jmasa, brutisso, coleenp, tschatzl

duke@435 1 /*
johnc@5277 2 * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
duke@435 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
duke@435 4 *
duke@435 5 * This code is free software; you can redistribute it and/or modify it
duke@435 6 * under the terms of the GNU General Public License version 2 only, as
duke@435 7 * published by the Free Software Foundation.
duke@435 8 *
duke@435 9 * This code is distributed in the hope that it will be useful, but WITHOUT
duke@435 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
duke@435 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
duke@435 12 * version 2 for more details (a copy is included in the LICENSE file that
duke@435 13 * accompanied this code).
duke@435 14 *
duke@435 15 * You should have received a copy of the GNU General Public License version
duke@435 16 * 2 along with this work; if not, write to the Free Software Foundation,
duke@435 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
duke@435 18 *
trims@1907 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
trims@1907 20 * or visit www.oracle.com if you need additional information or have any
trims@1907 21 * questions.
duke@435 22 *
duke@435 23 */
duke@435 24
stefank@2314 25 #ifndef SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP
stefank@2314 26 #define SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP
stefank@2314 27
stefank@2314 28 #include "memory/allocation.inline.hpp"
coleenp@2497 29 #include "oops/symbol.hpp"
stefank@2314 30 #include "utilities/hashtable.hpp"
stefank@2314 31
coleenp@2497 32 // The symbol table holds all Symbol*s and corresponding interned strings.
coleenp@2497 33 // Symbol*s and literal strings should be canonicalized.
duke@435 34 //
duke@435 35 // The interned strings are created lazily.
duke@435 36 //
duke@435 37 // It is implemented as an open hash table with a fixed number of buckets.
duke@435 38 //
duke@435 39 // %note:
duke@435 40 // - symbolTableEntrys are allocated in blocks to reduce the space overhead.
duke@435 41
duke@435 42 class BoolObjectClosure;
coleenp@3865 43 class outputStream;
duke@435 44
duke@435 45
coleenp@2497 46 // Class to hold a newly created or referenced Symbol* temporarily in scope.
coleenp@2497 47 // new_symbol() and lookup() will create a Symbol* if not already in the
coleenp@2497 48 // symbol table and add to the symbol's reference count.
coleenp@2497 49 // probe() and lookup_only() will increment the refcount if symbol is found.
coleenp@2497 50 class TempNewSymbol : public StackObj {
coleenp@2497 51 Symbol* _temp;
coleenp@2497 52
coleenp@2497 53 public:
coleenp@2497 54 TempNewSymbol() : _temp(NULL) {}
coleenp@2497 55 // Creating or looking up a symbol increments the symbol's reference count
coleenp@2497 56 TempNewSymbol(Symbol *s) : _temp(s) {}
coleenp@2497 57
coleenp@2497 58 // Operator= increments reference count.
coleenp@2497 59 void operator=(const TempNewSymbol &s) {
twisti@3969 60 //clear(); //FIXME
coleenp@2497 61 _temp = s._temp;
coleenp@2497 62 if (_temp !=NULL) _temp->increment_refcount();
coleenp@2497 63 }
coleenp@2497 64
coleenp@2497 65 // Decrement reference counter so it can go away if it's unique
twisti@3969 66 void clear() { if (_temp != NULL) _temp->decrement_refcount(); _temp = NULL; }
twisti@3969 67
twisti@3969 68 ~TempNewSymbol() { clear(); }
coleenp@2497 69
coleenp@2497 70 // Operators so they can be used like Symbols
coleenp@2497 71 Symbol* operator -> () const { return _temp; }
coleenp@2497 72 bool operator == (Symbol* o) const { return _temp == o; }
coleenp@2497 73 // Sneaky conversion function
coleenp@2497 74 operator Symbol*() { return _temp; }
coleenp@2497 75 };
coleenp@2497 76
mgerdin@7207 77 class SymbolTable : public RehashableHashtable<Symbol*, mtSymbol> {
duke@435 78 friend class VMStructs;
coleenp@2497 79 friend class ClassFileParser;
duke@435 80
duke@435 81 private:
duke@435 82 // The symbol table
duke@435 83 static SymbolTable* _the_table;
duke@435 84
coleenp@3865 85 // Set if one bucket is out of balance due to hash algorithm deficiency
coleenp@3865 86 static bool _needs_rehashing;
coleenp@3865 87
coleenp@2497 88 // For statistics
tschatzl@6229 89 static int _symbols_removed;
tschatzl@6229 90 static int _symbols_counted;
coleenp@2497 91
coleenp@3682 92 Symbol* allocate_symbol(const u1* name, int len, bool c_heap, TRAPS); // Assumes no characters larger than 0x7F
coleenp@2497 93
duke@435 94 // Adding elements
coleenp@3682 95 Symbol* basic_add(int index, u1* name, int len, unsigned int hashValue,
coleenp@3682 96 bool c_heap, TRAPS);
coleenp@4037 97 bool basic_add(ClassLoaderData* loader_data,
coleenp@4037 98 constantPoolHandle cp, int names_count,
duke@435 99 const char** names, int* lengths, int* cp_indices,
duke@435 100 unsigned int* hashValues, TRAPS);
duke@435 101
coleenp@4037 102 static void new_symbols(ClassLoaderData* loader_data,
coleenp@4037 103 constantPoolHandle cp, int names_count,
coleenp@2497 104 const char** name, int* lengths,
coleenp@2497 105 int* cp_indices, unsigned int* hashValues,
coleenp@2497 106 TRAPS) {
coleenp@4037 107 add(loader_data, cp, names_count, name, lengths, cp_indices, hashValues, THREAD);
coleenp@2497 108 }
coleenp@2497 109
coleenp@2497 110 Symbol* lookup(int index, const char* name, int len, unsigned int hash);
duke@435 111
duke@435 112 SymbolTable()
mgerdin@7207 113 : RehashableHashtable<Symbol*, mtSymbol>(SymbolTableSize, sizeof (HashtableEntry<Symbol*, mtSymbol>)) {}
duke@435 114
zgu@3900 115 SymbolTable(HashtableBucket<mtSymbol>* t, int number_of_entries)
mgerdin@7207 116 : RehashableHashtable<Symbol*, mtSymbol>(SymbolTableSize, sizeof (HashtableEntry<Symbol*, mtSymbol>), t,
duke@435 117 number_of_entries) {}
duke@435 118
coleenp@3682 119 // Arena for permanent symbols (null class loader) that are never unloaded
coleenp@3682 120 static Arena* _arena;
coleenp@3682 121 static Arena* arena() { return _arena; } // called for statistics
duke@435 122
coleenp@3682 123 static void initialize_symbols(int arena_alloc_size = 0);
tschatzl@6229 124
tschatzl@6229 125 static volatile int _parallel_claimed_idx;
tschatzl@6229 126
tschatzl@6229 127 // Release any dead symbols
tschatzl@6229 128 static void buckets_unlink(int start_idx, int end_idx, int* processed, int* removed, size_t* memory_total);
duke@435 129 public:
duke@435 130 enum {
coleenp@3682 131 symbol_alloc_batch_size = 8,
coleenp@3682 132 // Pick initial size based on java -version size measurements
coleenp@3682 133 symbol_alloc_arena_size = 360*K
duke@435 134 };
duke@435 135
duke@435 136 // The symbol table
duke@435 137 static SymbolTable* the_table() { return _the_table; }
duke@435 138
kevinw@5850 139 // Size of one bucket in the string table. Used when checking for rollover.
kevinw@5850 140 static uint bucket_size() { return sizeof(HashtableBucket<mtSymbol>); }
kevinw@5850 141
duke@435 142 static void create_table() {
duke@435 143 assert(_the_table == NULL, "One symbol table allowed.");
duke@435 144 _the_table = new SymbolTable();
coleenp@3682 145 initialize_symbols(symbol_alloc_arena_size);
duke@435 146 }
duke@435 147
zgu@3900 148 static void create_table(HashtableBucket<mtSymbol>* t, int length,
duke@435 149 int number_of_entries) {
duke@435 150 assert(_the_table == NULL, "One symbol table allowed.");
kevinw@5850 151
kevinw@5850 152 // If CDS archive used a different symbol table size, use that size instead
kevinw@5850 153 // which is better than giving an error.
kevinw@5850 154 SymbolTableSize = length/bucket_size();
kevinw@5850 155
duke@435 156 _the_table = new SymbolTable(t, number_of_entries);
coleenp@3682 157 // if CDS give symbol table a default arena size since most symbols
coleenp@3682 158 // are already allocated in the shared misc section.
coleenp@3682 159 initialize_symbols();
duke@435 160 }
duke@435 161
coleenp@3875 162 static unsigned int hash_symbol(const char* s, int len);
coleenp@3865 163
coleenp@2497 164 static Symbol* lookup(const char* name, int len, TRAPS);
duke@435 165 // lookup only, won't add. Also calculate hash.
coleenp@2497 166 static Symbol* lookup_only(const char* name, int len, unsigned int& hash);
duke@435 167 // Only copy to C string to be added if lookup failed.
coleenp@2497 168 static Symbol* lookup(const Symbol* sym, int begin, int end, TRAPS);
coleenp@2497 169
coleenp@2497 170 static void release(Symbol* sym);
duke@435 171
phh@3427 172 // Look up the address of the literal in the SymbolTable for this Symbol*
phh@3427 173 static Symbol** lookup_symbol_addr(Symbol* sym);
phh@3427 174
jrose@1100 175 // jchar (utf16) version of lookups
coleenp@2497 176 static Symbol* lookup_unicode(const jchar* name, int len, TRAPS);
coleenp@2497 177 static Symbol* lookup_only_unicode(const jchar* name, int len, unsigned int& hash);
jrose@1100 178
coleenp@4037 179 static void add(ClassLoaderData* loader_data,
coleenp@4037 180 constantPoolHandle cp, int names_count,
duke@435 181 const char** names, int* lengths, int* cp_indices,
duke@435 182 unsigned int* hashValues, TRAPS);
duke@435 183
coleenp@2497 184 // Release any dead symbols
tschatzl@6229 185 static void unlink() {
tschatzl@6229 186 int processed = 0;
tschatzl@6229 187 int removed = 0;
tschatzl@6229 188 unlink(&processed, &removed);
tschatzl@6229 189 }
tschatzl@6229 190 static void unlink(int* processed, int* removed);
tschatzl@6229 191 // Release any dead symbols, possibly parallel version
tschatzl@6229 192 static void possibly_parallel_unlink(int* processed, int* removed);
coleenp@2497 193
coleenp@2497 194 // iterate over symbols
coleenp@2497 195 static void symbols_do(SymbolClosure *cl);
coleenp@2497 196
coleenp@2497 197 // Symbol creation
coleenp@2497 198 static Symbol* new_symbol(const char* utf8_buffer, int length, TRAPS) {
coleenp@2497 199 assert(utf8_buffer != NULL, "just checking");
coleenp@2497 200 return lookup(utf8_buffer, length, THREAD);
duke@435 201 }
coleenp@2497 202 static Symbol* new_symbol(const char* name, TRAPS) {
coleenp@2497 203 return new_symbol(name, (int)strlen(name), THREAD);
coleenp@2497 204 }
coleenp@2497 205 static Symbol* new_symbol(const Symbol* sym, int begin, int end, TRAPS) {
coleenp@2497 206 assert(begin <= end && end <= sym->utf8_length(), "just checking");
coleenp@2497 207 return lookup(sym, begin, end, THREAD);
duke@435 208 }
duke@435 209
coleenp@3682 210 // Create a symbol in the arena for symbols that are not deleted
coleenp@3682 211 static Symbol* new_permanent_symbol(const char* name, TRAPS);
coleenp@3682 212
duke@435 213 // Symbol lookup
coleenp@2497 214 static Symbol* lookup(int index, const char* name, int len, TRAPS);
duke@435 215
duke@435 216 // Needed for preloading classes in signatures when compiling.
duke@435 217 // Returns the symbol is already present in symbol table, otherwise
duke@435 218 // NULL. NO ALLOCATION IS GUARANTEED!
coleenp@2497 219 static Symbol* probe(const char* name, int len) {
jrose@1100 220 unsigned int ignore_hash;
jrose@1100 221 return lookup_only(name, len, ignore_hash);
jrose@1100 222 }
coleenp@2497 223 static Symbol* probe_unicode(const jchar* name, int len) {
jrose@1100 224 unsigned int ignore_hash;
jrose@1100 225 return lookup_only_unicode(name, len, ignore_hash);
jrose@1100 226 }
duke@435 227
duke@435 228 // Histogram
duke@435 229 static void print_histogram() PRODUCT_RETURN;
coleenp@2497 230 static void print() PRODUCT_RETURN;
duke@435 231
duke@435 232 // Debugging
duke@435 233 static void verify();
coleenp@3865 234 static void dump(outputStream* st);
duke@435 235
duke@435 236 // Sharing
duke@435 237 static void copy_buckets(char** top, char*end) {
zgu@3900 238 the_table()->Hashtable<Symbol*, mtSymbol>::copy_buckets(top, end);
duke@435 239 }
duke@435 240 static void copy_table(char** top, char*end) {
zgu@3900 241 the_table()->Hashtable<Symbol*, mtSymbol>::copy_table(top, end);
duke@435 242 }
duke@435 243 static void reverse(void* boundary = NULL) {
zgu@3900 244 the_table()->Hashtable<Symbol*, mtSymbol>::reverse(boundary);
duke@435 245 }
coleenp@3865 246
coleenp@3865 247 // Rehash the symbol table if it gets out of balance
coleenp@3865 248 static void rehash_table();
coleenp@3865 249 static bool needs_rehashing() { return _needs_rehashing; }
tschatzl@6229 250 // Parallel chunked scanning
tschatzl@6229 251 static void clear_parallel_claimed_index() { _parallel_claimed_idx = 0; }
tschatzl@6229 252 static int parallel_claimed_index() { return _parallel_claimed_idx; }
duke@435 253 };
duke@435 254
mgerdin@7207 255 class StringTable : public RehashableHashtable<oop, mtSymbol> {
duke@435 256 friend class VMStructs;
duke@435 257
duke@435 258 private:
duke@435 259 // The string table
duke@435 260 static StringTable* _the_table;
duke@435 261
coleenp@3865 262 // Set if one bucket is out of balance due to hash algorithm deficiency
coleenp@3865 263 static bool _needs_rehashing;
coleenp@3865 264
johnc@5277 265 // Claimed high water mark for parallel chunked scanning
johnc@5277 266 static volatile int _parallel_claimed_idx;
johnc@5277 267
duke@435 268 static oop intern(Handle string_or_null, jchar* chars, int length, TRAPS);
duke@435 269 oop basic_add(int index, Handle string_or_null, jchar* name, int len,
duke@435 270 unsigned int hashValue, TRAPS);
duke@435 271
duke@435 272 oop lookup(int index, jchar* chars, int length, unsigned int hashValue);
duke@435 273
johnc@5277 274 // Apply the give oop closure to the entries to the buckets
johnc@5277 275 // in the range [start_idx, end_idx).
tschatzl@6229 276 static void buckets_oops_do(OopClosure* f, int start_idx, int end_idx);
tschatzl@6229 277 // Unlink or apply the give oop closure to the entries to the buckets
tschatzl@6229 278 // in the range [start_idx, end_idx).
tschatzl@6229 279 static void buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClosure* f, int start_idx, int end_idx, int* processed, int* removed);
johnc@5277 280
mgerdin@7207 281 StringTable() : RehashableHashtable<oop, mtSymbol>((int)StringTableSize,
zgu@3900 282 sizeof (HashtableEntry<oop, mtSymbol>)) {}
duke@435 283
zgu@3900 284 StringTable(HashtableBucket<mtSymbol>* t, int number_of_entries)
mgerdin@7207 285 : RehashableHashtable<oop, mtSymbol>((int)StringTableSize, sizeof (HashtableEntry<oop, mtSymbol>), t,
jcoomes@2660 286 number_of_entries) {}
duke@435 287 public:
duke@435 288 // The string table
duke@435 289 static StringTable* the_table() { return _the_table; }
duke@435 290
hseigel@4277 291 // Size of one bucket in the string table. Used when checking for rollover.
hseigel@4277 292 static uint bucket_size() { return sizeof(HashtableBucket<mtSymbol>); }
hseigel@4277 293
duke@435 294 static void create_table() {
duke@435 295 assert(_the_table == NULL, "One string table allowed.");
duke@435 296 _the_table = new StringTable();
duke@435 297 }
duke@435 298
duke@435 299 // GC support
duke@435 300 // Delete pointers to otherwise-unreachable objects.
tschatzl@6229 301 static void unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f) {
tschatzl@6229 302 int processed = 0;
tschatzl@6229 303 int removed = 0;
tschatzl@6229 304 unlink_or_oops_do(cl, f, &processed, &removed);
tschatzl@6229 305 }
stefank@5196 306 static void unlink(BoolObjectClosure* cl) {
tschatzl@6229 307 int processed = 0;
tschatzl@6229 308 int removed = 0;
tschatzl@6229 309 unlink_or_oops_do(cl, NULL, &processed, &removed);
stefank@5196 310 }
tschatzl@6229 311 static void unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f, int* processed, int* removed);
tschatzl@6229 312 static void unlink(BoolObjectClosure* cl, int* processed, int* removed) {
tschatzl@6229 313 unlink_or_oops_do(cl, NULL, processed, removed);
tschatzl@6229 314 }
johnc@5277 315 // Serially invoke "f->do_oop" on the locations of all oops in the table.
coleenp@2497 316 static void oops_do(OopClosure* f);
duke@435 317
tschatzl@6229 318 // Possibly parallel versions of the above
tschatzl@6229 319 static void possibly_parallel_unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f, int* processed, int* removed);
tschatzl@6229 320 static void possibly_parallel_unlink(BoolObjectClosure* cl, int* processed, int* removed) {
tschatzl@6229 321 possibly_parallel_unlink_or_oops_do(cl, NULL, processed, removed);
tschatzl@6229 322 }
johnc@5277 323 static void possibly_parallel_oops_do(OopClosure* f);
johnc@5277 324
coleenp@3865 325 // Hashing algorithm, used as the hash value used by the
coleenp@3865 326 // StringTable for bucket selection and comparison (stored in the
coleenp@3865 327 // HashtableEntry structures). This is used in the String.intern() method.
coleenp@3875 328 static unsigned int hash_string(const jchar* s, int len);
coleenp@3865 329
coleenp@3865 330 // Internal test.
coleenp@3865 331 static void test_alt_hash() PRODUCT_RETURN;
coleenp@3865 332
duke@435 333 // Probing
coleenp@2497 334 static oop lookup(Symbol* symbol);
mgerdin@4850 335 static oop lookup(jchar* chars, int length);
duke@435 336
duke@435 337 // Interning
coleenp@2497 338 static oop intern(Symbol* symbol, TRAPS);
duke@435 339 static oop intern(oop string, TRAPS);
duke@435 340 static oop intern(const char *utf8_string, TRAPS);
duke@435 341
duke@435 342 // Debugging
duke@435 343 static void verify();
coleenp@3865 344 static void dump(outputStream* st);
duke@435 345
dcubed@5743 346 enum VerifyMesgModes {
dcubed@5743 347 _verify_quietly = 0,
dcubed@5743 348 _verify_with_mesgs = 1
dcubed@5743 349 };
dcubed@5743 350
dcubed@5743 351 enum VerifyRetTypes {
dcubed@5743 352 _verify_pass = 0,
dcubed@5743 353 _verify_fail_continue = 1,
dcubed@5743 354 _verify_fail_done = 2
dcubed@5743 355 };
dcubed@5743 356
dcubed@5743 357 static VerifyRetTypes compare_entries(int bkt1, int e_cnt1,
dcubed@5743 358 HashtableEntry<oop, mtSymbol>* e_ptr1,
dcubed@5743 359 int bkt2, int e_cnt2,
dcubed@5743 360 HashtableEntry<oop, mtSymbol>* e_ptr2);
dcubed@5743 361 static VerifyRetTypes verify_entry(int bkt, int e_cnt,
dcubed@5743 362 HashtableEntry<oop, mtSymbol>* e_ptr,
dcubed@5743 363 VerifyMesgModes mesg_mode);
dcubed@5743 364 static int verify_and_compare_entries();
dcubed@5743 365
duke@435 366 // Sharing
duke@435 367 static void copy_buckets(char** top, char*end) {
zgu@3900 368 the_table()->Hashtable<oop, mtSymbol>::copy_buckets(top, end);
duke@435 369 }
duke@435 370 static void copy_table(char** top, char*end) {
zgu@3900 371 the_table()->Hashtable<oop, mtSymbol>::copy_table(top, end);
duke@435 372 }
duke@435 373 static void reverse() {
zgu@3900 374 the_table()->Hashtable<oop, mtSymbol>::reverse();
duke@435 375 }
coleenp@3865 376
coleenp@3865 377 // Rehash the symbol table if it gets out of balance
coleenp@3865 378 static void rehash_table();
coleenp@3865 379 static bool needs_rehashing() { return _needs_rehashing; }
johnc@5277 380
johnc@5277 381 // Parallel chunked scanning
johnc@5277 382 static void clear_parallel_claimed_index() { _parallel_claimed_idx = 0; }
tschatzl@6229 383 static int parallel_claimed_index() { return _parallel_claimed_idx; }
duke@435 384 };
stefank@2314 385 #endif // SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP

mercurial