Thu, 27 Jan 2011 16:11:27 -0800
6990754: Use native memory and reference counting to implement SymbolTable
Summary: move symbols from permgen into C heap and reference count them
Reviewed-by: never, acorn, jmasa, stefank
duke@435 | 1 | /* |
stefank@2314 | 2 | * Copyright (c) 1997, 2010, 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 | #include "precompiled.hpp" |
stefank@2314 | 26 | #include "classfile/javaClasses.hpp" |
stefank@2314 | 27 | #include "classfile/symbolTable.hpp" |
stefank@2314 | 28 | #include "classfile/systemDictionary.hpp" |
stefank@2314 | 29 | #include "gc_interface/collectedHeap.inline.hpp" |
stefank@2314 | 30 | #include "memory/filemap.hpp" |
stefank@2314 | 31 | #include "memory/gcLocker.inline.hpp" |
stefank@2314 | 32 | #include "oops/oop.inline.hpp" |
stefank@2314 | 33 | #include "oops/oop.inline2.hpp" |
stefank@2314 | 34 | #include "runtime/mutexLocker.hpp" |
stefank@2314 | 35 | #include "utilities/hashtable.inline.hpp" |
duke@435 | 36 | |
duke@435 | 37 | // -------------------------------------------------------------------------- |
duke@435 | 38 | |
duke@435 | 39 | SymbolTable* SymbolTable::_the_table = NULL; |
duke@435 | 40 | |
coleenp@2497 | 41 | Symbol* SymbolTable::allocate_symbol(const u1* name, int len, TRAPS) { |
coleenp@2497 | 42 | // Don't allow symbols to be created which cannot fit in a Symbol*. |
coleenp@2497 | 43 | if (len > Symbol::max_length()) { |
coleenp@2497 | 44 | THROW_MSG_0(vmSymbols::java_lang_InternalError(), |
coleenp@2497 | 45 | "name is too long to represent"); |
coleenp@2497 | 46 | } |
coleenp@2497 | 47 | Symbol* sym = new (len) Symbol(name, len); |
coleenp@2497 | 48 | assert(sym != NULL, "new should call vm_exit_out_of_memory if C_HEAP is exhausted"); |
coleenp@2497 | 49 | return sym; |
coleenp@2497 | 50 | } |
coleenp@2497 | 51 | |
coleenp@2497 | 52 | bool SymbolTable::allocate_symbols(int names_count, const u1** names, |
coleenp@2497 | 53 | int* lengths, Symbol** syms, TRAPS) { |
coleenp@2497 | 54 | for (int i = 0; i< names_count; i++) { |
coleenp@2497 | 55 | if (lengths[i] > Symbol::max_length()) { |
coleenp@2497 | 56 | THROW_MSG_0(vmSymbols::java_lang_InternalError(), |
coleenp@2497 | 57 | "name is too long to represent"); |
coleenp@2497 | 58 | } |
coleenp@2497 | 59 | } |
coleenp@2497 | 60 | |
coleenp@2497 | 61 | for (int i = 0; i< names_count; i++) { |
coleenp@2497 | 62 | int len = lengths[i]; |
coleenp@2497 | 63 | syms[i] = new (len) Symbol(names[i], len); |
coleenp@2497 | 64 | assert(syms[i] != NULL, "new should call vm_exit_out_of_memory if " |
coleenp@2497 | 65 | "C_HEAP is exhausted"); |
coleenp@2497 | 66 | } |
coleenp@2497 | 67 | return true; |
coleenp@2497 | 68 | } |
coleenp@2497 | 69 | |
coleenp@2497 | 70 | // Call function for all symbols in the symbol table. |
coleenp@2497 | 71 | void SymbolTable::symbols_do(SymbolClosure *cl) { |
coleenp@2497 | 72 | const int n = the_table()->table_size(); |
coleenp@2497 | 73 | for (int i = 0; i < n; i++) { |
coleenp@2497 | 74 | for (HashtableEntry<Symbol*>* p = the_table()->bucket(i); |
coleenp@2497 | 75 | p != NULL; |
coleenp@2497 | 76 | p = p->next()) { |
coleenp@2497 | 77 | cl->do_symbol(p->literal_addr()); |
coleenp@2497 | 78 | } |
coleenp@2497 | 79 | } |
coleenp@2497 | 80 | } |
coleenp@2497 | 81 | |
coleenp@2497 | 82 | int SymbolTable::symbols_removed = 0; |
coleenp@2497 | 83 | int SymbolTable::symbols_counted = 0; |
coleenp@2497 | 84 | |
coleenp@2497 | 85 | // Remove unreferenced symbols from the symbol table |
coleenp@2497 | 86 | // This is done late during GC. This doesn't use the hash table unlink because |
coleenp@2497 | 87 | // it assumes that the literals are oops. |
coleenp@2497 | 88 | void SymbolTable::unlink() { |
coleenp@2497 | 89 | int removed = 0; |
coleenp@2497 | 90 | int total = 0; |
coleenp@2497 | 91 | int memory_total = 0; |
coleenp@2497 | 92 | for (int i = 0; i < the_table()->table_size(); ++i) { |
coleenp@2497 | 93 | for (HashtableEntry<Symbol*>** p = the_table()->bucket_addr(i); *p != NULL; ) { |
coleenp@2497 | 94 | HashtableEntry<Symbol*>* entry = *p; |
coleenp@2497 | 95 | if (entry->is_shared()) { |
coleenp@2497 | 96 | break; |
coleenp@2497 | 97 | } |
coleenp@2497 | 98 | Symbol* s = entry->literal(); |
coleenp@2497 | 99 | memory_total += s->object_size(); |
coleenp@2497 | 100 | total++; |
coleenp@2497 | 101 | assert(s != NULL, "just checking"); |
coleenp@2497 | 102 | // If reference count is zero, remove. |
coleenp@2497 | 103 | if (s->refcount() == 0) { |
coleenp@2497 | 104 | delete s; |
coleenp@2497 | 105 | removed++; |
coleenp@2497 | 106 | *p = entry->next(); |
coleenp@2497 | 107 | the_table()->free_entry(entry); |
coleenp@2497 | 108 | } else { |
coleenp@2497 | 109 | p = entry->next_addr(); |
coleenp@2497 | 110 | } |
coleenp@2497 | 111 | } |
coleenp@2497 | 112 | } |
coleenp@2497 | 113 | symbols_removed += removed; |
coleenp@2497 | 114 | symbols_counted += total; |
coleenp@2497 | 115 | if (PrintGCDetails) { |
coleenp@2497 | 116 | gclog_or_tty->print(" [Symbols=%d size=%dK] ", total, |
coleenp@2497 | 117 | (memory_total*HeapWordSize)/1024); |
coleenp@2497 | 118 | } |
coleenp@2497 | 119 | } |
coleenp@2497 | 120 | |
coleenp@2497 | 121 | |
duke@435 | 122 | // Lookup a symbol in a bucket. |
duke@435 | 123 | |
coleenp@2497 | 124 | Symbol* SymbolTable::lookup(int index, const char* name, |
duke@435 | 125 | int len, unsigned int hash) { |
coleenp@2497 | 126 | for (HashtableEntry<Symbol*>* e = bucket(index); e != NULL; e = e->next()) { |
duke@435 | 127 | if (e->hash() == hash) { |
coleenp@2497 | 128 | Symbol* sym = e->literal(); |
duke@435 | 129 | if (sym->equals(name, len)) { |
coleenp@2497 | 130 | // something is referencing this symbol now. |
coleenp@2497 | 131 | sym->increment_refcount(); |
duke@435 | 132 | return sym; |
duke@435 | 133 | } |
duke@435 | 134 | } |
duke@435 | 135 | } |
duke@435 | 136 | return NULL; |
duke@435 | 137 | } |
duke@435 | 138 | |
duke@435 | 139 | |
duke@435 | 140 | // We take care not to be blocking while holding the |
duke@435 | 141 | // SymbolTable_lock. Otherwise, the system might deadlock, since the |
duke@435 | 142 | // symboltable is used during compilation (VM_thread) The lock free |
duke@435 | 143 | // synchronization is simplified by the fact that we do not delete |
duke@435 | 144 | // entries in the symbol table during normal execution (only during |
duke@435 | 145 | // safepoints). |
duke@435 | 146 | |
coleenp@2497 | 147 | Symbol* SymbolTable::lookup(const char* name, int len, TRAPS) { |
duke@435 | 148 | unsigned int hashValue = hash_symbol(name, len); |
duke@435 | 149 | int index = the_table()->hash_to_index(hashValue); |
duke@435 | 150 | |
coleenp@2497 | 151 | Symbol* s = the_table()->lookup(index, name, len, hashValue); |
duke@435 | 152 | |
duke@435 | 153 | // Found |
duke@435 | 154 | if (s != NULL) return s; |
duke@435 | 155 | |
duke@435 | 156 | // Otherwise, add to symbol to table |
duke@435 | 157 | return the_table()->basic_add(index, (u1*)name, len, hashValue, CHECK_NULL); |
duke@435 | 158 | } |
duke@435 | 159 | |
coleenp@2497 | 160 | Symbol* SymbolTable::lookup(const Symbol* sym, int begin, int end, TRAPS) { |
duke@435 | 161 | char* buffer; |
duke@435 | 162 | int index, len; |
duke@435 | 163 | unsigned int hashValue; |
duke@435 | 164 | char* name; |
duke@435 | 165 | { |
duke@435 | 166 | debug_only(No_Safepoint_Verifier nsv;) |
duke@435 | 167 | |
duke@435 | 168 | name = (char*)sym->base() + begin; |
duke@435 | 169 | len = end - begin; |
duke@435 | 170 | hashValue = hash_symbol(name, len); |
duke@435 | 171 | index = the_table()->hash_to_index(hashValue); |
coleenp@2497 | 172 | Symbol* s = the_table()->lookup(index, name, len, hashValue); |
duke@435 | 173 | |
duke@435 | 174 | // Found |
duke@435 | 175 | if (s != NULL) return s; |
duke@435 | 176 | } |
duke@435 | 177 | |
duke@435 | 178 | // Otherwise, add to symbol to table. Copy to a C string first. |
duke@435 | 179 | char stack_buf[128]; |
duke@435 | 180 | ResourceMark rm(THREAD); |
duke@435 | 181 | if (len <= 128) { |
duke@435 | 182 | buffer = stack_buf; |
duke@435 | 183 | } else { |
duke@435 | 184 | buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len); |
duke@435 | 185 | } |
duke@435 | 186 | for (int i=0; i<len; i++) { |
duke@435 | 187 | buffer[i] = name[i]; |
duke@435 | 188 | } |
duke@435 | 189 | // Make sure there is no safepoint in the code above since name can't move. |
duke@435 | 190 | // We can't include the code in No_Safepoint_Verifier because of the |
duke@435 | 191 | // ResourceMark. |
duke@435 | 192 | |
duke@435 | 193 | return the_table()->basic_add(index, (u1*)buffer, len, hashValue, CHECK_NULL); |
duke@435 | 194 | } |
duke@435 | 195 | |
coleenp@2497 | 196 | Symbol* SymbolTable::lookup_only(const char* name, int len, |
duke@435 | 197 | unsigned int& hash) { |
duke@435 | 198 | hash = hash_symbol(name, len); |
duke@435 | 199 | int index = the_table()->hash_to_index(hash); |
duke@435 | 200 | |
coleenp@2497 | 201 | Symbol* s = the_table()->lookup(index, name, len, hash); |
coleenp@2497 | 202 | return s; |
duke@435 | 203 | } |
duke@435 | 204 | |
jrose@1100 | 205 | // Suggestion: Push unicode-based lookup all the way into the hashing |
jrose@1100 | 206 | // and probing logic, so there is no need for convert_to_utf8 until |
coleenp@2497 | 207 | // an actual new Symbol* is created. |
coleenp@2497 | 208 | Symbol* SymbolTable::lookup_unicode(const jchar* name, int utf16_length, TRAPS) { |
jrose@1100 | 209 | int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length); |
jrose@1100 | 210 | char stack_buf[128]; |
jrose@1100 | 211 | if (utf8_length < (int) sizeof(stack_buf)) { |
jrose@1100 | 212 | char* chars = stack_buf; |
jrose@1100 | 213 | UNICODE::convert_to_utf8(name, utf16_length, chars); |
jrose@1100 | 214 | return lookup(chars, utf8_length, THREAD); |
jrose@1100 | 215 | } else { |
jrose@1100 | 216 | ResourceMark rm(THREAD); |
jrose@1100 | 217 | char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);; |
jrose@1100 | 218 | UNICODE::convert_to_utf8(name, utf16_length, chars); |
jrose@1100 | 219 | return lookup(chars, utf8_length, THREAD); |
jrose@1100 | 220 | } |
jrose@1100 | 221 | } |
jrose@1100 | 222 | |
coleenp@2497 | 223 | Symbol* SymbolTable::lookup_only_unicode(const jchar* name, int utf16_length, |
jrose@1100 | 224 | unsigned int& hash) { |
jrose@1100 | 225 | int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length); |
jrose@1100 | 226 | char stack_buf[128]; |
jrose@1100 | 227 | if (utf8_length < (int) sizeof(stack_buf)) { |
jrose@1100 | 228 | char* chars = stack_buf; |
jrose@1100 | 229 | UNICODE::convert_to_utf8(name, utf16_length, chars); |
jrose@1100 | 230 | return lookup_only(chars, utf8_length, hash); |
jrose@1100 | 231 | } else { |
jrose@1100 | 232 | ResourceMark rm; |
jrose@1100 | 233 | char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);; |
jrose@1100 | 234 | UNICODE::convert_to_utf8(name, utf16_length, chars); |
jrose@1100 | 235 | return lookup_only(chars, utf8_length, hash); |
jrose@1100 | 236 | } |
jrose@1100 | 237 | } |
jrose@1100 | 238 | |
duke@435 | 239 | void SymbolTable::add(constantPoolHandle cp, int names_count, |
duke@435 | 240 | const char** names, int* lengths, int* cp_indices, |
duke@435 | 241 | unsigned int* hashValues, TRAPS) { |
duke@435 | 242 | SymbolTable* table = the_table(); |
duke@435 | 243 | bool added = table->basic_add(cp, names_count, names, lengths, |
duke@435 | 244 | cp_indices, hashValues, CHECK); |
duke@435 | 245 | if (!added) { |
duke@435 | 246 | // do it the hard way |
duke@435 | 247 | for (int i=0; i<names_count; i++) { |
duke@435 | 248 | int index = table->hash_to_index(hashValues[i]); |
coleenp@2497 | 249 | Symbol* sym = table->basic_add(index, (u1*)names[i], lengths[i], |
duke@435 | 250 | hashValues[i], CHECK); |
duke@435 | 251 | cp->symbol_at_put(cp_indices[i], sym); |
duke@435 | 252 | } |
duke@435 | 253 | } |
duke@435 | 254 | } |
duke@435 | 255 | |
coleenp@2497 | 256 | Symbol* SymbolTable::basic_add(int index, u1 *name, int len, |
duke@435 | 257 | unsigned int hashValue, TRAPS) { |
duke@435 | 258 | assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(), |
duke@435 | 259 | "proposed name of symbol must be stable"); |
duke@435 | 260 | |
duke@435 | 261 | // We assume that lookup() has been called already, that it failed, |
duke@435 | 262 | // and symbol was not found. We create the symbol here. |
coleenp@2497 | 263 | Symbol* sym = allocate_symbol(name, len, CHECK_NULL); |
duke@435 | 264 | |
coleenp@2497 | 265 | // Allocation must be done before grabbing the SymbolTable_lock lock |
duke@435 | 266 | MutexLocker ml(SymbolTable_lock, THREAD); |
duke@435 | 267 | |
duke@435 | 268 | assert(sym->equals((char*)name, len), "symbol must be properly initialized"); |
duke@435 | 269 | |
duke@435 | 270 | // Since look-up was done lock-free, we need to check if another |
duke@435 | 271 | // thread beat us in the race to insert the symbol. |
duke@435 | 272 | |
coleenp@2497 | 273 | Symbol* test = lookup(index, (char*)name, len, hashValue); |
duke@435 | 274 | if (test != NULL) { |
twisti@1040 | 275 | // A race occurred and another thread introduced the symbol, this one |
duke@435 | 276 | // will be dropped and collected. |
coleenp@2497 | 277 | delete sym; |
coleenp@2497 | 278 | assert(test->refcount() != 0, "lookup should have incremented the count"); |
duke@435 | 279 | return test; |
duke@435 | 280 | } |
duke@435 | 281 | |
coleenp@2497 | 282 | HashtableEntry<Symbol*>* entry = new_entry(hashValue, sym); |
coleenp@2497 | 283 | sym->increment_refcount(); |
duke@435 | 284 | add_entry(index, entry); |
coleenp@2497 | 285 | return sym; |
duke@435 | 286 | } |
duke@435 | 287 | |
duke@435 | 288 | bool SymbolTable::basic_add(constantPoolHandle cp, int names_count, |
duke@435 | 289 | const char** names, int* lengths, |
duke@435 | 290 | int* cp_indices, unsigned int* hashValues, |
duke@435 | 291 | TRAPS) { |
coleenp@2497 | 292 | Symbol* syms[symbol_alloc_batch_size]; |
coleenp@2497 | 293 | bool allocated = allocate_symbols(names_count, (const u1**)names, lengths, |
coleenp@2497 | 294 | syms, CHECK_false); |
duke@435 | 295 | if (!allocated) { |
duke@435 | 296 | return false; |
duke@435 | 297 | } |
duke@435 | 298 | |
duke@435 | 299 | // Allocation must be done before grabbing the SymbolTable_lock lock |
duke@435 | 300 | MutexLocker ml(SymbolTable_lock, THREAD); |
duke@435 | 301 | |
coleenp@2497 | 302 | for (int i=0; i<names_count; i++) { |
duke@435 | 303 | assert(syms[i]->equals(names[i], lengths[i]), "symbol must be properly initialized"); |
duke@435 | 304 | // Since look-up was done lock-free, we need to check if another |
duke@435 | 305 | // thread beat us in the race to insert the symbol. |
duke@435 | 306 | int index = hash_to_index(hashValues[i]); |
coleenp@2497 | 307 | Symbol* test = lookup(index, names[i], lengths[i], hashValues[i]); |
duke@435 | 308 | if (test != NULL) { |
twisti@1040 | 309 | // A race occurred and another thread introduced the symbol, this one |
duke@435 | 310 | // will be dropped and collected. Use test instead. |
duke@435 | 311 | cp->symbol_at_put(cp_indices[i], test); |
coleenp@2497 | 312 | assert(test->refcount() != 0, "lookup should have incremented the count"); |
coleenp@2497 | 313 | delete syms[i]; |
duke@435 | 314 | } else { |
coleenp@2497 | 315 | Symbol* sym = syms[i]; |
coleenp@2497 | 316 | HashtableEntry<Symbol*>* entry = new_entry(hashValues[i], sym); |
coleenp@2497 | 317 | sym->increment_refcount(); // increment refcount in external hashtable |
duke@435 | 318 | add_entry(index, entry); |
duke@435 | 319 | cp->symbol_at_put(cp_indices[i], sym); |
duke@435 | 320 | } |
duke@435 | 321 | } |
duke@435 | 322 | |
duke@435 | 323 | return true; |
duke@435 | 324 | } |
duke@435 | 325 | |
duke@435 | 326 | |
duke@435 | 327 | void SymbolTable::verify() { |
duke@435 | 328 | for (int i = 0; i < the_table()->table_size(); ++i) { |
coleenp@2497 | 329 | HashtableEntry<Symbol*>* p = the_table()->bucket(i); |
duke@435 | 330 | for ( ; p != NULL; p = p->next()) { |
coleenp@2497 | 331 | Symbol* s = (Symbol*)(p->literal()); |
duke@435 | 332 | guarantee(s != NULL, "symbol is NULL"); |
duke@435 | 333 | unsigned int h = hash_symbol((char*)s->bytes(), s->utf8_length()); |
duke@435 | 334 | guarantee(p->hash() == h, "broken hash in symbol table entry"); |
duke@435 | 335 | guarantee(the_table()->hash_to_index(h) == i, |
duke@435 | 336 | "wrong index in symbol table"); |
duke@435 | 337 | } |
duke@435 | 338 | } |
duke@435 | 339 | } |
duke@435 | 340 | |
duke@435 | 341 | |
duke@435 | 342 | //--------------------------------------------------------------------------- |
duke@435 | 343 | // Non-product code |
duke@435 | 344 | |
duke@435 | 345 | #ifndef PRODUCT |
duke@435 | 346 | |
duke@435 | 347 | void SymbolTable::print_histogram() { |
duke@435 | 348 | MutexLocker ml(SymbolTable_lock); |
duke@435 | 349 | const int results_length = 100; |
duke@435 | 350 | int results[results_length]; |
duke@435 | 351 | int i,j; |
duke@435 | 352 | |
duke@435 | 353 | // initialize results to zero |
duke@435 | 354 | for (j = 0; j < results_length; j++) { |
duke@435 | 355 | results[j] = 0; |
duke@435 | 356 | } |
duke@435 | 357 | |
duke@435 | 358 | int total = 0; |
duke@435 | 359 | int max_symbols = 0; |
duke@435 | 360 | int out_of_range = 0; |
coleenp@2497 | 361 | int memory_total = 0; |
coleenp@2497 | 362 | int count = 0; |
duke@435 | 363 | for (i = 0; i < the_table()->table_size(); i++) { |
coleenp@2497 | 364 | HashtableEntry<Symbol*>* p = the_table()->bucket(i); |
duke@435 | 365 | for ( ; p != NULL; p = p->next()) { |
coleenp@2497 | 366 | memory_total += p->literal()->object_size(); |
coleenp@2497 | 367 | count++; |
coleenp@2497 | 368 | int counter = p->literal()->utf8_length(); |
duke@435 | 369 | total += counter; |
duke@435 | 370 | if (counter < results_length) { |
duke@435 | 371 | results[counter]++; |
duke@435 | 372 | } else { |
duke@435 | 373 | out_of_range++; |
duke@435 | 374 | } |
duke@435 | 375 | max_symbols = MAX2(max_symbols, counter); |
duke@435 | 376 | } |
duke@435 | 377 | } |
duke@435 | 378 | tty->print_cr("Symbol Table:"); |
coleenp@2497 | 379 | tty->print_cr("Total number of symbols %5d", count); |
coleenp@2497 | 380 | tty->print_cr("Total size in memory %5dK", |
coleenp@2497 | 381 | (memory_total*HeapWordSize)/1024); |
coleenp@2497 | 382 | tty->print_cr("Total counted %5d", symbols_counted); |
coleenp@2497 | 383 | tty->print_cr("Total removed %5d", symbols_removed); |
coleenp@2497 | 384 | if (symbols_counted > 0) { |
coleenp@2497 | 385 | tty->print_cr("Percent removed %3.2f", |
coleenp@2497 | 386 | ((float)symbols_removed/(float)symbols_counted)* 100); |
coleenp@2497 | 387 | } |
coleenp@2497 | 388 | tty->print_cr("Reference counts %5d", Symbol::_total_count); |
coleenp@2497 | 389 | tty->print_cr("Histogram of symbol length:"); |
duke@435 | 390 | tty->print_cr("%8s %5d", "Total ", total); |
duke@435 | 391 | tty->print_cr("%8s %5d", "Maximum", max_symbols); |
duke@435 | 392 | tty->print_cr("%8s %3.2f", "Average", |
duke@435 | 393 | ((float) total / (float) the_table()->table_size())); |
duke@435 | 394 | tty->print_cr("%s", "Histogram:"); |
duke@435 | 395 | tty->print_cr(" %s %29s", "Length", "Number chains that length"); |
duke@435 | 396 | for (i = 0; i < results_length; i++) { |
duke@435 | 397 | if (results[i] > 0) { |
duke@435 | 398 | tty->print_cr("%6d %10d", i, results[i]); |
duke@435 | 399 | } |
duke@435 | 400 | } |
coleenp@2497 | 401 | if (Verbose) { |
coleenp@2497 | 402 | int line_length = 70; |
coleenp@2497 | 403 | tty->print_cr("%s %30s", " Length", "Number chains that length"); |
coleenp@2497 | 404 | for (i = 0; i < results_length; i++) { |
coleenp@2497 | 405 | if (results[i] > 0) { |
coleenp@2497 | 406 | tty->print("%4d", i); |
coleenp@2497 | 407 | for (j = 0; (j < results[i]) && (j < line_length); j++) { |
coleenp@2497 | 408 | tty->print("%1s", "*"); |
coleenp@2497 | 409 | } |
coleenp@2497 | 410 | if (j == line_length) { |
coleenp@2497 | 411 | tty->print("%1s", "+"); |
coleenp@2497 | 412 | } |
coleenp@2497 | 413 | tty->cr(); |
duke@435 | 414 | } |
coleenp@2497 | 415 | } |
coleenp@2497 | 416 | } |
coleenp@2497 | 417 | tty->print_cr(" %s %d: %d\n", "Number chains longer than", |
coleenp@2497 | 418 | results_length, out_of_range); |
coleenp@2497 | 419 | } |
coleenp@2497 | 420 | |
coleenp@2497 | 421 | void SymbolTable::print() { |
coleenp@2497 | 422 | for (int i = 0; i < the_table()->table_size(); ++i) { |
coleenp@2497 | 423 | HashtableEntry<Symbol*>** p = the_table()->bucket_addr(i); |
coleenp@2497 | 424 | HashtableEntry<Symbol*>* entry = the_table()->bucket(i); |
coleenp@2497 | 425 | if (entry != NULL) { |
coleenp@2497 | 426 | while (entry != NULL) { |
coleenp@2497 | 427 | tty->print(PTR_FORMAT " ", entry->literal()); |
coleenp@2497 | 428 | entry->literal()->print(); |
coleenp@2497 | 429 | tty->print(" %d", entry->literal()->refcount()); |
coleenp@2497 | 430 | p = entry->next_addr(); |
coleenp@2497 | 431 | entry = (HashtableEntry<Symbol*>*)HashtableEntry<Symbol*>::make_ptr(*p); |
duke@435 | 432 | } |
duke@435 | 433 | tty->cr(); |
duke@435 | 434 | } |
duke@435 | 435 | } |
duke@435 | 436 | } |
duke@435 | 437 | |
duke@435 | 438 | #endif // PRODUCT |
duke@435 | 439 | |
duke@435 | 440 | // -------------------------------------------------------------------------- |
duke@435 | 441 | |
duke@435 | 442 | #ifdef ASSERT |
duke@435 | 443 | class StableMemoryChecker : public StackObj { |
duke@435 | 444 | enum { _bufsize = wordSize*4 }; |
duke@435 | 445 | |
duke@435 | 446 | address _region; |
duke@435 | 447 | jint _size; |
duke@435 | 448 | u1 _save_buf[_bufsize]; |
duke@435 | 449 | |
duke@435 | 450 | int sample(u1* save_buf) { |
duke@435 | 451 | if (_size <= _bufsize) { |
duke@435 | 452 | memcpy(save_buf, _region, _size); |
duke@435 | 453 | return _size; |
duke@435 | 454 | } else { |
duke@435 | 455 | // copy head and tail |
duke@435 | 456 | memcpy(&save_buf[0], _region, _bufsize/2); |
duke@435 | 457 | memcpy(&save_buf[_bufsize/2], _region + _size - _bufsize/2, _bufsize/2); |
duke@435 | 458 | return (_bufsize/2)*2; |
duke@435 | 459 | } |
duke@435 | 460 | } |
duke@435 | 461 | |
duke@435 | 462 | public: |
duke@435 | 463 | StableMemoryChecker(const void* region, jint size) { |
duke@435 | 464 | _region = (address) region; |
duke@435 | 465 | _size = size; |
duke@435 | 466 | sample(_save_buf); |
duke@435 | 467 | } |
duke@435 | 468 | |
duke@435 | 469 | bool verify() { |
duke@435 | 470 | u1 check_buf[sizeof(_save_buf)]; |
duke@435 | 471 | int check_size = sample(check_buf); |
duke@435 | 472 | return (0 == memcmp(_save_buf, check_buf, check_size)); |
duke@435 | 473 | } |
duke@435 | 474 | |
duke@435 | 475 | void set_region(const void* region) { _region = (address) region; } |
duke@435 | 476 | }; |
duke@435 | 477 | #endif |
duke@435 | 478 | |
duke@435 | 479 | |
duke@435 | 480 | // -------------------------------------------------------------------------- |
duke@435 | 481 | |
duke@435 | 482 | |
duke@435 | 483 | // Compute the hash value for a java.lang.String object which would |
duke@435 | 484 | // contain the characters passed in. This hash value is used for at |
duke@435 | 485 | // least two purposes. |
duke@435 | 486 | // |
duke@435 | 487 | // (a) As the hash value used by the StringTable for bucket selection |
duke@435 | 488 | // and comparison (stored in the HashtableEntry structures). This |
duke@435 | 489 | // is used in the String.intern() method. |
duke@435 | 490 | // |
duke@435 | 491 | // (b) As the hash value used by the String object itself, in |
duke@435 | 492 | // String.hashCode(). This value is normally calculate in Java code |
duke@435 | 493 | // in the String.hashCode method(), but is precomputed for String |
duke@435 | 494 | // objects in the shared archive file. |
duke@435 | 495 | // |
duke@435 | 496 | // For this reason, THIS ALGORITHM MUST MATCH String.hashCode(). |
duke@435 | 497 | |
duke@435 | 498 | int StringTable::hash_string(jchar* s, int len) { |
duke@435 | 499 | unsigned h = 0; |
duke@435 | 500 | while (len-- > 0) { |
duke@435 | 501 | h = 31*h + (unsigned) *s; |
duke@435 | 502 | s++; |
duke@435 | 503 | } |
duke@435 | 504 | return h; |
duke@435 | 505 | } |
duke@435 | 506 | |
duke@435 | 507 | |
duke@435 | 508 | StringTable* StringTable::_the_table = NULL; |
duke@435 | 509 | |
duke@435 | 510 | oop StringTable::lookup(int index, jchar* name, |
duke@435 | 511 | int len, unsigned int hash) { |
coleenp@2497 | 512 | for (HashtableEntry<oop>* l = bucket(index); l != NULL; l = l->next()) { |
duke@435 | 513 | if (l->hash() == hash) { |
duke@435 | 514 | if (java_lang_String::equals(l->literal(), name, len)) { |
duke@435 | 515 | return l->literal(); |
duke@435 | 516 | } |
duke@435 | 517 | } |
duke@435 | 518 | } |
duke@435 | 519 | return NULL; |
duke@435 | 520 | } |
duke@435 | 521 | |
duke@435 | 522 | |
duke@435 | 523 | oop StringTable::basic_add(int index, Handle string_or_null, jchar* name, |
duke@435 | 524 | int len, unsigned int hashValue, TRAPS) { |
duke@435 | 525 | debug_only(StableMemoryChecker smc(name, len * sizeof(name[0]))); |
duke@435 | 526 | assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(), |
duke@435 | 527 | "proposed name of symbol must be stable"); |
duke@435 | 528 | |
duke@435 | 529 | Handle string; |
duke@435 | 530 | // try to reuse the string if possible |
duke@435 | 531 | if (!string_or_null.is_null() && string_or_null()->is_perm()) { |
duke@435 | 532 | string = string_or_null; |
duke@435 | 533 | } else { |
duke@435 | 534 | string = java_lang_String::create_tenured_from_unicode(name, len, CHECK_NULL); |
duke@435 | 535 | } |
duke@435 | 536 | |
duke@435 | 537 | // Allocation must be done before grapping the SymbolTable_lock lock |
duke@435 | 538 | MutexLocker ml(StringTable_lock, THREAD); |
duke@435 | 539 | |
duke@435 | 540 | assert(java_lang_String::equals(string(), name, len), |
duke@435 | 541 | "string must be properly initialized"); |
duke@435 | 542 | |
duke@435 | 543 | // Since look-up was done lock-free, we need to check if another |
duke@435 | 544 | // thread beat us in the race to insert the symbol. |
duke@435 | 545 | |
duke@435 | 546 | oop test = lookup(index, name, len, hashValue); // calls lookup(u1*, int) |
duke@435 | 547 | if (test != NULL) { |
duke@435 | 548 | // Entry already added |
duke@435 | 549 | return test; |
duke@435 | 550 | } |
duke@435 | 551 | |
coleenp@2497 | 552 | HashtableEntry<oop>* entry = new_entry(hashValue, string()); |
duke@435 | 553 | add_entry(index, entry); |
duke@435 | 554 | return string(); |
duke@435 | 555 | } |
duke@435 | 556 | |
duke@435 | 557 | |
coleenp@2497 | 558 | oop StringTable::lookup(Symbol* symbol) { |
duke@435 | 559 | ResourceMark rm; |
duke@435 | 560 | int length; |
duke@435 | 561 | jchar* chars = symbol->as_unicode(length); |
duke@435 | 562 | unsigned int hashValue = hash_string(chars, length); |
duke@435 | 563 | int index = the_table()->hash_to_index(hashValue); |
duke@435 | 564 | return the_table()->lookup(index, chars, length, hashValue); |
duke@435 | 565 | } |
duke@435 | 566 | |
duke@435 | 567 | |
duke@435 | 568 | oop StringTable::intern(Handle string_or_null, jchar* name, |
duke@435 | 569 | int len, TRAPS) { |
duke@435 | 570 | unsigned int hashValue = hash_string(name, len); |
duke@435 | 571 | int index = the_table()->hash_to_index(hashValue); |
duke@435 | 572 | oop string = the_table()->lookup(index, name, len, hashValue); |
duke@435 | 573 | |
duke@435 | 574 | // Found |
duke@435 | 575 | if (string != NULL) return string; |
duke@435 | 576 | |
duke@435 | 577 | // Otherwise, add to symbol to table |
duke@435 | 578 | return the_table()->basic_add(index, string_or_null, name, len, |
duke@435 | 579 | hashValue, CHECK_NULL); |
duke@435 | 580 | } |
duke@435 | 581 | |
coleenp@2497 | 582 | oop StringTable::intern(Symbol* symbol, TRAPS) { |
duke@435 | 583 | if (symbol == NULL) return NULL; |
duke@435 | 584 | ResourceMark rm(THREAD); |
duke@435 | 585 | int length; |
duke@435 | 586 | jchar* chars = symbol->as_unicode(length); |
duke@435 | 587 | Handle string; |
duke@435 | 588 | oop result = intern(string, chars, length, CHECK_NULL); |
duke@435 | 589 | return result; |
duke@435 | 590 | } |
duke@435 | 591 | |
duke@435 | 592 | |
duke@435 | 593 | oop StringTable::intern(oop string, TRAPS) |
duke@435 | 594 | { |
duke@435 | 595 | if (string == NULL) return NULL; |
duke@435 | 596 | ResourceMark rm(THREAD); |
duke@435 | 597 | int length; |
duke@435 | 598 | Handle h_string (THREAD, string); |
duke@435 | 599 | jchar* chars = java_lang_String::as_unicode_string(string, length); |
duke@435 | 600 | oop result = intern(h_string, chars, length, CHECK_NULL); |
duke@435 | 601 | return result; |
duke@435 | 602 | } |
duke@435 | 603 | |
duke@435 | 604 | |
duke@435 | 605 | oop StringTable::intern(const char* utf8_string, TRAPS) { |
duke@435 | 606 | if (utf8_string == NULL) return NULL; |
duke@435 | 607 | ResourceMark rm(THREAD); |
duke@435 | 608 | int length = UTF8::unicode_length(utf8_string); |
duke@435 | 609 | jchar* chars = NEW_RESOURCE_ARRAY(jchar, length); |
duke@435 | 610 | UTF8::convert_to_unicode(utf8_string, chars, length); |
duke@435 | 611 | Handle string; |
duke@435 | 612 | oop result = intern(string, chars, length, CHECK_NULL); |
duke@435 | 613 | return result; |
duke@435 | 614 | } |
duke@435 | 615 | |
coleenp@2497 | 616 | void StringTable::unlink(BoolObjectClosure* is_alive) { |
coleenp@2497 | 617 | // Readers of the table are unlocked, so we should only be removing |
coleenp@2497 | 618 | // entries at a safepoint. |
coleenp@2497 | 619 | assert(SafepointSynchronize::is_at_safepoint(), "must be at safepoint"); |
coleenp@2497 | 620 | for (int i = 0; i < the_table()->table_size(); ++i) { |
coleenp@2497 | 621 | for (HashtableEntry<oop>** p = the_table()->bucket_addr(i); *p != NULL; ) { |
coleenp@2497 | 622 | HashtableEntry<oop>* entry = *p; |
coleenp@2497 | 623 | if (entry->is_shared()) { |
coleenp@2497 | 624 | break; |
coleenp@2497 | 625 | } |
coleenp@2497 | 626 | assert(entry->literal() != NULL, "just checking"); |
coleenp@2497 | 627 | if (is_alive->do_object_b(entry->literal())) { |
coleenp@2497 | 628 | p = entry->next_addr(); |
coleenp@2497 | 629 | } else { |
coleenp@2497 | 630 | *p = entry->next(); |
coleenp@2497 | 631 | the_table()->free_entry(entry); |
coleenp@2497 | 632 | } |
coleenp@2497 | 633 | } |
coleenp@2497 | 634 | } |
coleenp@2497 | 635 | } |
coleenp@2497 | 636 | |
coleenp@2497 | 637 | void StringTable::oops_do(OopClosure* f) { |
coleenp@2497 | 638 | for (int i = 0; i < the_table()->table_size(); ++i) { |
coleenp@2497 | 639 | HashtableEntry<oop>** p = the_table()->bucket_addr(i); |
coleenp@2497 | 640 | HashtableEntry<oop>* entry = the_table()->bucket(i); |
coleenp@2497 | 641 | while (entry != NULL) { |
coleenp@2497 | 642 | f->do_oop((oop*)entry->literal_addr()); |
coleenp@2497 | 643 | |
coleenp@2497 | 644 | // Did the closure remove the literal from the table? |
coleenp@2497 | 645 | if (entry->literal() == NULL) { |
coleenp@2497 | 646 | assert(!entry->is_shared(), "immutable hashtable entry?"); |
coleenp@2497 | 647 | *p = entry->next(); |
coleenp@2497 | 648 | the_table()->free_entry(entry); |
coleenp@2497 | 649 | } else { |
coleenp@2497 | 650 | p = entry->next_addr(); |
coleenp@2497 | 651 | } |
coleenp@2497 | 652 | entry = (HashtableEntry<oop>*)HashtableEntry<oop>::make_ptr(*p); |
coleenp@2497 | 653 | } |
coleenp@2497 | 654 | } |
coleenp@2497 | 655 | } |
coleenp@2497 | 656 | |
duke@435 | 657 | void StringTable::verify() { |
duke@435 | 658 | for (int i = 0; i < the_table()->table_size(); ++i) { |
coleenp@2497 | 659 | HashtableEntry<oop>* p = the_table()->bucket(i); |
duke@435 | 660 | for ( ; p != NULL; p = p->next()) { |
duke@435 | 661 | oop s = p->literal(); |
duke@435 | 662 | guarantee(s != NULL, "interned string is NULL"); |
duke@435 | 663 | guarantee(s->is_perm(), "interned string not in permspace"); |
duke@435 | 664 | |
duke@435 | 665 | int length; |
duke@435 | 666 | jchar* chars = java_lang_String::as_unicode_string(s, length); |
duke@435 | 667 | unsigned int h = hash_string(chars, length); |
duke@435 | 668 | guarantee(p->hash() == h, "broken hash in string table entry"); |
duke@435 | 669 | guarantee(the_table()->hash_to_index(h) == i, |
duke@435 | 670 | "wrong index in string table"); |
duke@435 | 671 | } |
duke@435 | 672 | } |
duke@435 | 673 | } |