1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/src/share/vm/classfile/symbolTable.cpp Sat Dec 01 00:00:00 2007 +0000
1.3 @@ -0,0 +1,484 @@
1.4 +/*
1.5 + * Copyright 1997-2006 Sun Microsystems, Inc. All Rights Reserved.
1.6 + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
1.7 + *
1.8 + * This code is free software; you can redistribute it and/or modify it
1.9 + * under the terms of the GNU General Public License version 2 only, as
1.10 + * published by the Free Software Foundation.
1.11 + *
1.12 + * This code is distributed in the hope that it will be useful, but WITHOUT
1.13 + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
1.14 + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
1.15 + * version 2 for more details (a copy is included in the LICENSE file that
1.16 + * accompanied this code).
1.17 + *
1.18 + * You should have received a copy of the GNU General Public License version
1.19 + * 2 along with this work; if not, write to the Free Software Foundation,
1.20 + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
1.21 + *
1.22 + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
1.23 + * CA 95054 USA or visit www.sun.com if you need additional information or
1.24 + * have any questions.
1.25 + *
1.26 + */
1.27 +
1.28 +# include "incls/_precompiled.incl"
1.29 +# include "incls/_symbolTable.cpp.incl"
1.30 +
1.31 +// --------------------------------------------------------------------------
1.32 +
1.33 +SymbolTable* SymbolTable::_the_table = NULL;
1.34 +
1.35 +// Lookup a symbol in a bucket.
1.36 +
1.37 +symbolOop SymbolTable::lookup(int index, const char* name,
1.38 + int len, unsigned int hash) {
1.39 + for (HashtableEntry* e = bucket(index); e != NULL; e = e->next()) {
1.40 + if (e->hash() == hash) {
1.41 + symbolOop sym = symbolOop(e->literal());
1.42 + if (sym->equals(name, len)) {
1.43 + return sym;
1.44 + }
1.45 + }
1.46 + }
1.47 + return NULL;
1.48 +}
1.49 +
1.50 +
1.51 +// We take care not to be blocking while holding the
1.52 +// SymbolTable_lock. Otherwise, the system might deadlock, since the
1.53 +// symboltable is used during compilation (VM_thread) The lock free
1.54 +// synchronization is simplified by the fact that we do not delete
1.55 +// entries in the symbol table during normal execution (only during
1.56 +// safepoints).
1.57 +
1.58 +symbolOop SymbolTable::lookup(const char* name, int len, TRAPS) {
1.59 + unsigned int hashValue = hash_symbol(name, len);
1.60 + int index = the_table()->hash_to_index(hashValue);
1.61 +
1.62 + symbolOop s = the_table()->lookup(index, name, len, hashValue);
1.63 +
1.64 + // Found
1.65 + if (s != NULL) return s;
1.66 +
1.67 + // Otherwise, add to symbol to table
1.68 + return the_table()->basic_add(index, (u1*)name, len, hashValue, CHECK_NULL);
1.69 +}
1.70 +
1.71 +symbolOop SymbolTable::lookup(symbolHandle sym, int begin, int end, TRAPS) {
1.72 + char* buffer;
1.73 + int index, len;
1.74 + unsigned int hashValue;
1.75 + char* name;
1.76 + {
1.77 + debug_only(No_Safepoint_Verifier nsv;)
1.78 +
1.79 + name = (char*)sym->base() + begin;
1.80 + len = end - begin;
1.81 + hashValue = hash_symbol(name, len);
1.82 + index = the_table()->hash_to_index(hashValue);
1.83 + symbolOop s = the_table()->lookup(index, name, len, hashValue);
1.84 +
1.85 + // Found
1.86 + if (s != NULL) return s;
1.87 + }
1.88 +
1.89 + // Otherwise, add to symbol to table. Copy to a C string first.
1.90 + char stack_buf[128];
1.91 + ResourceMark rm(THREAD);
1.92 + if (len <= 128) {
1.93 + buffer = stack_buf;
1.94 + } else {
1.95 + buffer = NEW_RESOURCE_ARRAY_IN_THREAD(THREAD, char, len);
1.96 + }
1.97 + for (int i=0; i<len; i++) {
1.98 + buffer[i] = name[i];
1.99 + }
1.100 + // Make sure there is no safepoint in the code above since name can't move.
1.101 + // We can't include the code in No_Safepoint_Verifier because of the
1.102 + // ResourceMark.
1.103 +
1.104 + return the_table()->basic_add(index, (u1*)buffer, len, hashValue, CHECK_NULL);
1.105 +}
1.106 +
1.107 +symbolOop SymbolTable::lookup_only(const char* name, int len,
1.108 + unsigned int& hash) {
1.109 + hash = hash_symbol(name, len);
1.110 + int index = the_table()->hash_to_index(hash);
1.111 +
1.112 + return the_table()->lookup(index, name, len, hash);
1.113 +}
1.114 +
1.115 +void SymbolTable::add(constantPoolHandle cp, int names_count,
1.116 + const char** names, int* lengths, int* cp_indices,
1.117 + unsigned int* hashValues, TRAPS) {
1.118 + SymbolTable* table = the_table();
1.119 + bool added = table->basic_add(cp, names_count, names, lengths,
1.120 + cp_indices, hashValues, CHECK);
1.121 + if (!added) {
1.122 + // do it the hard way
1.123 + for (int i=0; i<names_count; i++) {
1.124 + int index = table->hash_to_index(hashValues[i]);
1.125 + symbolOop sym = table->basic_add(index, (u1*)names[i], lengths[i],
1.126 + hashValues[i], CHECK);
1.127 + cp->symbol_at_put(cp_indices[i], sym);
1.128 + }
1.129 + }
1.130 +}
1.131 +
1.132 +// Needed for preloading classes in signatures when compiling.
1.133 +
1.134 +symbolOop SymbolTable::probe(const char* name, int len) {
1.135 + unsigned int hashValue = hash_symbol(name, len);
1.136 + int index = the_table()->hash_to_index(hashValue);
1.137 + return the_table()->lookup(index, name, len, hashValue);
1.138 +}
1.139 +
1.140 +
1.141 +symbolOop SymbolTable::basic_add(int index, u1 *name, int len,
1.142 + unsigned int hashValue, TRAPS) {
1.143 + assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),
1.144 + "proposed name of symbol must be stable");
1.145 +
1.146 + // We assume that lookup() has been called already, that it failed,
1.147 + // and symbol was not found. We create the symbol here.
1.148 + symbolKlass* sk = (symbolKlass*) Universe::symbolKlassObj()->klass_part();
1.149 + symbolOop s_oop = sk->allocate_symbol(name, len, CHECK_NULL);
1.150 + symbolHandle sym (THREAD, s_oop);
1.151 +
1.152 + // Allocation must be done before grapping the SymbolTable_lock lock
1.153 + MutexLocker ml(SymbolTable_lock, THREAD);
1.154 +
1.155 + assert(sym->equals((char*)name, len), "symbol must be properly initialized");
1.156 +
1.157 + // Since look-up was done lock-free, we need to check if another
1.158 + // thread beat us in the race to insert the symbol.
1.159 +
1.160 + symbolOop test = lookup(index, (char*)name, len, hashValue);
1.161 + if (test != NULL) {
1.162 + // A race occured and another thread introduced the symbol, this one
1.163 + // will be dropped and collected.
1.164 + return test;
1.165 + }
1.166 +
1.167 + HashtableEntry* entry = new_entry(hashValue, sym());
1.168 + add_entry(index, entry);
1.169 + return sym();
1.170 +}
1.171 +
1.172 +bool SymbolTable::basic_add(constantPoolHandle cp, int names_count,
1.173 + const char** names, int* lengths,
1.174 + int* cp_indices, unsigned int* hashValues,
1.175 + TRAPS) {
1.176 + symbolKlass* sk = (symbolKlass*) Universe::symbolKlassObj()->klass_part();
1.177 + symbolOop sym_oops[symbol_alloc_batch_size];
1.178 + bool allocated = sk->allocate_symbols(names_count, names, lengths,
1.179 + sym_oops, CHECK_false);
1.180 + if (!allocated) {
1.181 + return false;
1.182 + }
1.183 + symbolHandle syms[symbol_alloc_batch_size];
1.184 + int i;
1.185 + for (i=0; i<names_count; i++) {
1.186 + syms[i] = symbolHandle(THREAD, sym_oops[i]);
1.187 + }
1.188 +
1.189 + // Allocation must be done before grabbing the SymbolTable_lock lock
1.190 + MutexLocker ml(SymbolTable_lock, THREAD);
1.191 +
1.192 + for (i=0; i<names_count; i++) {
1.193 + assert(syms[i]->equals(names[i], lengths[i]), "symbol must be properly initialized");
1.194 + // Since look-up was done lock-free, we need to check if another
1.195 + // thread beat us in the race to insert the symbol.
1.196 + int index = hash_to_index(hashValues[i]);
1.197 + symbolOop test = lookup(index, names[i], lengths[i], hashValues[i]);
1.198 + if (test != NULL) {
1.199 + // A race occured and another thread introduced the symbol, this one
1.200 + // will be dropped and collected. Use test instead.
1.201 + cp->symbol_at_put(cp_indices[i], test);
1.202 + } else {
1.203 + symbolOop sym = syms[i]();
1.204 + HashtableEntry* entry = new_entry(hashValues[i], sym);
1.205 + add_entry(index, entry);
1.206 + cp->symbol_at_put(cp_indices[i], sym);
1.207 + }
1.208 + }
1.209 +
1.210 + return true;
1.211 +}
1.212 +
1.213 +
1.214 +void SymbolTable::verify() {
1.215 + for (int i = 0; i < the_table()->table_size(); ++i) {
1.216 + HashtableEntry* p = the_table()->bucket(i);
1.217 + for ( ; p != NULL; p = p->next()) {
1.218 + symbolOop s = symbolOop(p->literal());
1.219 + guarantee(s != NULL, "symbol is NULL");
1.220 + s->verify();
1.221 + guarantee(s->is_perm(), "symbol not in permspace");
1.222 + unsigned int h = hash_symbol((char*)s->bytes(), s->utf8_length());
1.223 + guarantee(p->hash() == h, "broken hash in symbol table entry");
1.224 + guarantee(the_table()->hash_to_index(h) == i,
1.225 + "wrong index in symbol table");
1.226 + }
1.227 + }
1.228 +}
1.229 +
1.230 +
1.231 +//---------------------------------------------------------------------------
1.232 +// Non-product code
1.233 +
1.234 +#ifndef PRODUCT
1.235 +
1.236 +void SymbolTable::print_histogram() {
1.237 + MutexLocker ml(SymbolTable_lock);
1.238 + const int results_length = 100;
1.239 + int results[results_length];
1.240 + int i,j;
1.241 +
1.242 + // initialize results to zero
1.243 + for (j = 0; j < results_length; j++) {
1.244 + results[j] = 0;
1.245 + }
1.246 +
1.247 + int total = 0;
1.248 + int max_symbols = 0;
1.249 + int out_of_range = 0;
1.250 + for (i = 0; i < the_table()->table_size(); i++) {
1.251 + HashtableEntry* p = the_table()->bucket(i);
1.252 + for ( ; p != NULL; p = p->next()) {
1.253 + int counter = symbolOop(p->literal())->utf8_length();
1.254 + total += counter;
1.255 + if (counter < results_length) {
1.256 + results[counter]++;
1.257 + } else {
1.258 + out_of_range++;
1.259 + }
1.260 + max_symbols = MAX2(max_symbols, counter);
1.261 + }
1.262 + }
1.263 + tty->print_cr("Symbol Table:");
1.264 + tty->print_cr("%8s %5d", "Total ", total);
1.265 + tty->print_cr("%8s %5d", "Maximum", max_symbols);
1.266 + tty->print_cr("%8s %3.2f", "Average",
1.267 + ((float) total / (float) the_table()->table_size()));
1.268 + tty->print_cr("%s", "Histogram:");
1.269 + tty->print_cr(" %s %29s", "Length", "Number chains that length");
1.270 + for (i = 0; i < results_length; i++) {
1.271 + if (results[i] > 0) {
1.272 + tty->print_cr("%6d %10d", i, results[i]);
1.273 + }
1.274 + }
1.275 + int line_length = 70;
1.276 + tty->print_cr("%s %30s", " Length", "Number chains that length");
1.277 + for (i = 0; i < results_length; i++) {
1.278 + if (results[i] > 0) {
1.279 + tty->print("%4d", i);
1.280 + for (j = 0; (j < results[i]) && (j < line_length); j++) {
1.281 + tty->print("%1s", "*");
1.282 + }
1.283 + if (j == line_length) {
1.284 + tty->print("%1s", "+");
1.285 + }
1.286 + tty->cr();
1.287 + }
1.288 + }
1.289 + tty->print_cr(" %s %d: %d\n", "Number chains longer than",
1.290 + results_length, out_of_range);
1.291 +}
1.292 +
1.293 +#endif // PRODUCT
1.294 +
1.295 +// --------------------------------------------------------------------------
1.296 +
1.297 +#ifdef ASSERT
1.298 +class StableMemoryChecker : public StackObj {
1.299 + enum { _bufsize = wordSize*4 };
1.300 +
1.301 + address _region;
1.302 + jint _size;
1.303 + u1 _save_buf[_bufsize];
1.304 +
1.305 + int sample(u1* save_buf) {
1.306 + if (_size <= _bufsize) {
1.307 + memcpy(save_buf, _region, _size);
1.308 + return _size;
1.309 + } else {
1.310 + // copy head and tail
1.311 + memcpy(&save_buf[0], _region, _bufsize/2);
1.312 + memcpy(&save_buf[_bufsize/2], _region + _size - _bufsize/2, _bufsize/2);
1.313 + return (_bufsize/2)*2;
1.314 + }
1.315 + }
1.316 +
1.317 + public:
1.318 + StableMemoryChecker(const void* region, jint size) {
1.319 + _region = (address) region;
1.320 + _size = size;
1.321 + sample(_save_buf);
1.322 + }
1.323 +
1.324 + bool verify() {
1.325 + u1 check_buf[sizeof(_save_buf)];
1.326 + int check_size = sample(check_buf);
1.327 + return (0 == memcmp(_save_buf, check_buf, check_size));
1.328 + }
1.329 +
1.330 + void set_region(const void* region) { _region = (address) region; }
1.331 +};
1.332 +#endif
1.333 +
1.334 +
1.335 +// --------------------------------------------------------------------------
1.336 +
1.337 +
1.338 +// Compute the hash value for a java.lang.String object which would
1.339 +// contain the characters passed in. This hash value is used for at
1.340 +// least two purposes.
1.341 +//
1.342 +// (a) As the hash value used by the StringTable for bucket selection
1.343 +// and comparison (stored in the HashtableEntry structures). This
1.344 +// is used in the String.intern() method.
1.345 +//
1.346 +// (b) As the hash value used by the String object itself, in
1.347 +// String.hashCode(). This value is normally calculate in Java code
1.348 +// in the String.hashCode method(), but is precomputed for String
1.349 +// objects in the shared archive file.
1.350 +//
1.351 +// For this reason, THIS ALGORITHM MUST MATCH String.hashCode().
1.352 +
1.353 +int StringTable::hash_string(jchar* s, int len) {
1.354 + unsigned h = 0;
1.355 + while (len-- > 0) {
1.356 + h = 31*h + (unsigned) *s;
1.357 + s++;
1.358 + }
1.359 + return h;
1.360 +}
1.361 +
1.362 +
1.363 +StringTable* StringTable::_the_table = NULL;
1.364 +
1.365 +oop StringTable::lookup(int index, jchar* name,
1.366 + int len, unsigned int hash) {
1.367 + for (HashtableEntry* l = bucket(index); l != NULL; l = l->next()) {
1.368 + if (l->hash() == hash) {
1.369 + if (java_lang_String::equals(l->literal(), name, len)) {
1.370 + return l->literal();
1.371 + }
1.372 + }
1.373 + }
1.374 + return NULL;
1.375 +}
1.376 +
1.377 +
1.378 +oop StringTable::basic_add(int index, Handle string_or_null, jchar* name,
1.379 + int len, unsigned int hashValue, TRAPS) {
1.380 + debug_only(StableMemoryChecker smc(name, len * sizeof(name[0])));
1.381 + assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),
1.382 + "proposed name of symbol must be stable");
1.383 +
1.384 + Handle string;
1.385 + // try to reuse the string if possible
1.386 + if (!string_or_null.is_null() && string_or_null()->is_perm()) {
1.387 + string = string_or_null;
1.388 + } else {
1.389 + string = java_lang_String::create_tenured_from_unicode(name, len, CHECK_NULL);
1.390 + }
1.391 +
1.392 + // Allocation must be done before grapping the SymbolTable_lock lock
1.393 + MutexLocker ml(StringTable_lock, THREAD);
1.394 +
1.395 + assert(java_lang_String::equals(string(), name, len),
1.396 + "string must be properly initialized");
1.397 +
1.398 + // Since look-up was done lock-free, we need to check if another
1.399 + // thread beat us in the race to insert the symbol.
1.400 +
1.401 + oop test = lookup(index, name, len, hashValue); // calls lookup(u1*, int)
1.402 + if (test != NULL) {
1.403 + // Entry already added
1.404 + return test;
1.405 + }
1.406 +
1.407 + HashtableEntry* entry = new_entry(hashValue, string());
1.408 + add_entry(index, entry);
1.409 + return string();
1.410 +}
1.411 +
1.412 +
1.413 +oop StringTable::lookup(symbolOop symbol) {
1.414 + ResourceMark rm;
1.415 + int length;
1.416 + jchar* chars = symbol->as_unicode(length);
1.417 + unsigned int hashValue = hash_string(chars, length);
1.418 + int index = the_table()->hash_to_index(hashValue);
1.419 + return the_table()->lookup(index, chars, length, hashValue);
1.420 +}
1.421 +
1.422 +
1.423 +oop StringTable::intern(Handle string_or_null, jchar* name,
1.424 + int len, TRAPS) {
1.425 + unsigned int hashValue = hash_string(name, len);
1.426 + int index = the_table()->hash_to_index(hashValue);
1.427 + oop string = the_table()->lookup(index, name, len, hashValue);
1.428 +
1.429 + // Found
1.430 + if (string != NULL) return string;
1.431 +
1.432 + // Otherwise, add to symbol to table
1.433 + return the_table()->basic_add(index, string_or_null, name, len,
1.434 + hashValue, CHECK_NULL);
1.435 +}
1.436 +
1.437 +oop StringTable::intern(symbolOop symbol, TRAPS) {
1.438 + if (symbol == NULL) return NULL;
1.439 + ResourceMark rm(THREAD);
1.440 + int length;
1.441 + jchar* chars = symbol->as_unicode(length);
1.442 + Handle string;
1.443 + oop result = intern(string, chars, length, CHECK_NULL);
1.444 + return result;
1.445 +}
1.446 +
1.447 +
1.448 +oop StringTable::intern(oop string, TRAPS)
1.449 +{
1.450 + if (string == NULL) return NULL;
1.451 + ResourceMark rm(THREAD);
1.452 + int length;
1.453 + Handle h_string (THREAD, string);
1.454 + jchar* chars = java_lang_String::as_unicode_string(string, length);
1.455 + oop result = intern(h_string, chars, length, CHECK_NULL);
1.456 + return result;
1.457 +}
1.458 +
1.459 +
1.460 +oop StringTable::intern(const char* utf8_string, TRAPS) {
1.461 + if (utf8_string == NULL) return NULL;
1.462 + ResourceMark rm(THREAD);
1.463 + int length = UTF8::unicode_length(utf8_string);
1.464 + jchar* chars = NEW_RESOURCE_ARRAY(jchar, length);
1.465 + UTF8::convert_to_unicode(utf8_string, chars, length);
1.466 + Handle string;
1.467 + oop result = intern(string, chars, length, CHECK_NULL);
1.468 + return result;
1.469 +}
1.470 +
1.471 +void StringTable::verify() {
1.472 + for (int i = 0; i < the_table()->table_size(); ++i) {
1.473 + HashtableEntry* p = the_table()->bucket(i);
1.474 + for ( ; p != NULL; p = p->next()) {
1.475 + oop s = p->literal();
1.476 + guarantee(s != NULL, "interned string is NULL");
1.477 + guarantee(s->is_perm(), "interned string not in permspace");
1.478 +
1.479 + int length;
1.480 + jchar* chars = java_lang_String::as_unicode_string(s, length);
1.481 + unsigned int h = hash_string(chars, length);
1.482 + guarantee(p->hash() == h, "broken hash in string table entry");
1.483 + guarantee(the_table()->hash_to_index(h) == i,
1.484 + "wrong index in string table");
1.485 + }
1.486 + }
1.487 +}
