src/share/vm/classfile/symbolTable.hpp

Tue, 18 Jun 2013 12:31:07 -0700

author
johnc
date
Tue, 18 Jun 2013 12:31:07 -0700
changeset 5277
01522ca68fc7
parent 5196
8dbc025ff709
child 5743
63147986a428
permissions
-rw-r--r--

8015237: Parallelize string table scanning during strong root processing
Summary: Parallelize the scanning of the intern string table by having each GC worker claim a given number of buckets. Changes were also reviewed by Per Liden <per.liden@oracle.com>.
Reviewed-by: tschatzl, stefank, twisti

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
zgu@3900 77 class SymbolTable : public Hashtable<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
coleenp@2497 89 static int symbols_removed;
coleenp@2497 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
duke@435 110 // Table size
duke@435 111 enum {
duke@435 112 symbol_table_size = 20011
duke@435 113 };
duke@435 114
coleenp@2497 115 Symbol* lookup(int index, const char* name, int len, unsigned int hash);
duke@435 116
duke@435 117 SymbolTable()
zgu@3900 118 : Hashtable<Symbol*, mtSymbol>(symbol_table_size, sizeof (HashtableEntry<Symbol*, mtSymbol>)) {}
duke@435 119
zgu@3900 120 SymbolTable(HashtableBucket<mtSymbol>* t, int number_of_entries)
zgu@3900 121 : Hashtable<Symbol*, mtSymbol>(symbol_table_size, sizeof (HashtableEntry<Symbol*, mtSymbol>), t,
duke@435 122 number_of_entries) {}
duke@435 123
coleenp@3682 124 // Arena for permanent symbols (null class loader) that are never unloaded
coleenp@3682 125 static Arena* _arena;
coleenp@3682 126 static Arena* arena() { return _arena; } // called for statistics
duke@435 127
coleenp@3682 128 static void initialize_symbols(int arena_alloc_size = 0);
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
duke@435 139 static void create_table() {
duke@435 140 assert(_the_table == NULL, "One symbol table allowed.");
duke@435 141 _the_table = new SymbolTable();
coleenp@3682 142 initialize_symbols(symbol_alloc_arena_size);
duke@435 143 }
duke@435 144
zgu@3900 145 static void create_table(HashtableBucket<mtSymbol>* t, int length,
duke@435 146 int number_of_entries) {
duke@435 147 assert(_the_table == NULL, "One symbol table allowed.");
zgu@3900 148 assert(length == symbol_table_size * sizeof(HashtableBucket<mtSymbol>),
duke@435 149 "bad shared symbol size.");
duke@435 150 _the_table = new SymbolTable(t, number_of_entries);
coleenp@3682 151 // if CDS give symbol table a default arena size since most symbols
coleenp@3682 152 // are already allocated in the shared misc section.
coleenp@3682 153 initialize_symbols();
duke@435 154 }
duke@435 155
coleenp@3875 156 static unsigned int hash_symbol(const char* s, int len);
coleenp@3865 157
coleenp@2497 158 static Symbol* lookup(const char* name, int len, TRAPS);
duke@435 159 // lookup only, won't add. Also calculate hash.
coleenp@2497 160 static Symbol* lookup_only(const char* name, int len, unsigned int& hash);
duke@435 161 // Only copy to C string to be added if lookup failed.
coleenp@2497 162 static Symbol* lookup(const Symbol* sym, int begin, int end, TRAPS);
coleenp@2497 163
coleenp@2497 164 static void release(Symbol* sym);
duke@435 165
phh@3427 166 // Look up the address of the literal in the SymbolTable for this Symbol*
phh@3427 167 static Symbol** lookup_symbol_addr(Symbol* sym);
phh@3427 168
jrose@1100 169 // jchar (utf16) version of lookups
coleenp@2497 170 static Symbol* lookup_unicode(const jchar* name, int len, TRAPS);
coleenp@2497 171 static Symbol* lookup_only_unicode(const jchar* name, int len, unsigned int& hash);
jrose@1100 172
coleenp@4037 173 static void add(ClassLoaderData* loader_data,
coleenp@4037 174 constantPoolHandle cp, int names_count,
duke@435 175 const char** names, int* lengths, int* cp_indices,
duke@435 176 unsigned int* hashValues, TRAPS);
duke@435 177
coleenp@2497 178 // Release any dead symbols
coleenp@2497 179 static void unlink();
coleenp@2497 180
coleenp@2497 181 // iterate over symbols
coleenp@2497 182 static void symbols_do(SymbolClosure *cl);
coleenp@2497 183
coleenp@2497 184 // Symbol creation
coleenp@2497 185 static Symbol* new_symbol(const char* utf8_buffer, int length, TRAPS) {
coleenp@2497 186 assert(utf8_buffer != NULL, "just checking");
coleenp@2497 187 return lookup(utf8_buffer, length, THREAD);
duke@435 188 }
coleenp@2497 189 static Symbol* new_symbol(const char* name, TRAPS) {
coleenp@2497 190 return new_symbol(name, (int)strlen(name), THREAD);
coleenp@2497 191 }
coleenp@2497 192 static Symbol* new_symbol(const Symbol* sym, int begin, int end, TRAPS) {
coleenp@2497 193 assert(begin <= end && end <= sym->utf8_length(), "just checking");
coleenp@2497 194 return lookup(sym, begin, end, THREAD);
duke@435 195 }
duke@435 196
coleenp@3682 197 // Create a symbol in the arena for symbols that are not deleted
coleenp@3682 198 static Symbol* new_permanent_symbol(const char* name, TRAPS);
coleenp@3682 199
duke@435 200 // Symbol lookup
coleenp@2497 201 static Symbol* lookup(int index, const char* name, int len, TRAPS);
duke@435 202
duke@435 203 // Needed for preloading classes in signatures when compiling.
duke@435 204 // Returns the symbol is already present in symbol table, otherwise
duke@435 205 // NULL. NO ALLOCATION IS GUARANTEED!
coleenp@2497 206 static Symbol* probe(const char* name, int len) {
jrose@1100 207 unsigned int ignore_hash;
jrose@1100 208 return lookup_only(name, len, ignore_hash);
jrose@1100 209 }
coleenp@2497 210 static Symbol* probe_unicode(const jchar* name, int len) {
jrose@1100 211 unsigned int ignore_hash;
jrose@1100 212 return lookup_only_unicode(name, len, ignore_hash);
jrose@1100 213 }
duke@435 214
duke@435 215 // Histogram
duke@435 216 static void print_histogram() PRODUCT_RETURN;
coleenp@2497 217 static void print() PRODUCT_RETURN;
duke@435 218
duke@435 219 // Debugging
duke@435 220 static void verify();
coleenp@3865 221 static void dump(outputStream* st);
duke@435 222
duke@435 223 // Sharing
duke@435 224 static void copy_buckets(char** top, char*end) {
zgu@3900 225 the_table()->Hashtable<Symbol*, mtSymbol>::copy_buckets(top, end);
duke@435 226 }
duke@435 227 static void copy_table(char** top, char*end) {
zgu@3900 228 the_table()->Hashtable<Symbol*, mtSymbol>::copy_table(top, end);
duke@435 229 }
duke@435 230 static void reverse(void* boundary = NULL) {
zgu@3900 231 the_table()->Hashtable<Symbol*, mtSymbol>::reverse(boundary);
duke@435 232 }
coleenp@3865 233
coleenp@3865 234 // Rehash the symbol table if it gets out of balance
coleenp@3865 235 static void rehash_table();
coleenp@3865 236 static bool needs_rehashing() { return _needs_rehashing; }
duke@435 237 };
duke@435 238
zgu@3900 239 class StringTable : public Hashtable<oop, mtSymbol> {
duke@435 240 friend class VMStructs;
duke@435 241
duke@435 242 private:
duke@435 243 // The string table
duke@435 244 static StringTable* _the_table;
duke@435 245
coleenp@3865 246 // Set if one bucket is out of balance due to hash algorithm deficiency
coleenp@3865 247 static bool _needs_rehashing;
coleenp@3865 248
johnc@5277 249 // Claimed high water mark for parallel chunked scanning
johnc@5277 250 static volatile int _parallel_claimed_idx;
johnc@5277 251
duke@435 252 static oop intern(Handle string_or_null, jchar* chars, int length, TRAPS);
duke@435 253 oop basic_add(int index, Handle string_or_null, jchar* name, int len,
duke@435 254 unsigned int hashValue, TRAPS);
duke@435 255
duke@435 256 oop lookup(int index, jchar* chars, int length, unsigned int hashValue);
duke@435 257
johnc@5277 258 // Apply the give oop closure to the entries to the buckets
johnc@5277 259 // in the range [start_idx, end_idx).
johnc@5277 260 static void buckets_do(OopClosure* f, int start_idx, int end_idx);
johnc@5277 261
zgu@3900 262 StringTable() : Hashtable<oop, mtSymbol>((int)StringTableSize,
zgu@3900 263 sizeof (HashtableEntry<oop, mtSymbol>)) {}
duke@435 264
zgu@3900 265 StringTable(HashtableBucket<mtSymbol>* t, int number_of_entries)
zgu@3900 266 : Hashtable<oop, mtSymbol>((int)StringTableSize, sizeof (HashtableEntry<oop, mtSymbol>), t,
jcoomes@2660 267 number_of_entries) {}
duke@435 268 public:
duke@435 269 // The string table
duke@435 270 static StringTable* the_table() { return _the_table; }
duke@435 271
hseigel@4277 272 // Size of one bucket in the string table. Used when checking for rollover.
hseigel@4277 273 static uint bucket_size() { return sizeof(HashtableBucket<mtSymbol>); }
hseigel@4277 274
duke@435 275 static void create_table() {
duke@435 276 assert(_the_table == NULL, "One string table allowed.");
duke@435 277 _the_table = new StringTable();
duke@435 278 }
duke@435 279
duke@435 280 // GC support
duke@435 281 // Delete pointers to otherwise-unreachable objects.
stefank@5196 282 static void unlink_or_oops_do(BoolObjectClosure* cl, OopClosure* f);
stefank@5196 283 static void unlink(BoolObjectClosure* cl) {
stefank@5196 284 unlink_or_oops_do(cl, NULL);
stefank@5196 285 }
duke@435 286
johnc@5277 287 // Serially invoke "f->do_oop" on the locations of all oops in the table.
coleenp@2497 288 static void oops_do(OopClosure* f);
duke@435 289
johnc@5277 290 // Possibly parallel version of the above
johnc@5277 291 static void possibly_parallel_oops_do(OopClosure* f);
johnc@5277 292
coleenp@3865 293 // Hashing algorithm, used as the hash value used by the
coleenp@3865 294 // StringTable for bucket selection and comparison (stored in the
coleenp@3865 295 // HashtableEntry structures). This is used in the String.intern() method.
coleenp@3875 296 static unsigned int hash_string(const jchar* s, int len);
coleenp@3865 297
coleenp@3865 298 // Internal test.
coleenp@3865 299 static void test_alt_hash() PRODUCT_RETURN;
coleenp@3865 300
duke@435 301 // Probing
coleenp@2497 302 static oop lookup(Symbol* symbol);
mgerdin@4850 303 static oop lookup(jchar* chars, int length);
duke@435 304
duke@435 305 // Interning
coleenp@2497 306 static oop intern(Symbol* symbol, TRAPS);
duke@435 307 static oop intern(oop string, TRAPS);
duke@435 308 static oop intern(const char *utf8_string, TRAPS);
duke@435 309
duke@435 310 // Debugging
duke@435 311 static void verify();
coleenp@3865 312 static void dump(outputStream* st);
duke@435 313
duke@435 314 // Sharing
duke@435 315 static void copy_buckets(char** top, char*end) {
zgu@3900 316 the_table()->Hashtable<oop, mtSymbol>::copy_buckets(top, end);
duke@435 317 }
duke@435 318 static void copy_table(char** top, char*end) {
zgu@3900 319 the_table()->Hashtable<oop, mtSymbol>::copy_table(top, end);
duke@435 320 }
duke@435 321 static void reverse() {
zgu@3900 322 the_table()->Hashtable<oop, mtSymbol>::reverse();
duke@435 323 }
coleenp@3865 324
coleenp@3865 325 // Rehash the symbol table if it gets out of balance
coleenp@3865 326 static void rehash_table();
coleenp@3865 327 static bool needs_rehashing() { return _needs_rehashing; }
johnc@5277 328
johnc@5277 329 // Parallel chunked scanning
johnc@5277 330 static void clear_parallel_claimed_index() { _parallel_claimed_idx = 0; }
duke@435 331 };
stefank@2314 332 #endif // SHARE_VM_CLASSFILE_SYMBOLTABLE_HPP

mercurial