Fri, 29 Aug 2014 13:08:01 +0200
8056084: Refactor Hashtable to allow implementations without rehashing support
Reviewed-by: gziemski, jmasa, brutisso, coleenp, tschatzl
1 /*
2 * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
25 #include "precompiled.hpp"
26 #include "classfile/altHashing.hpp"
27 #include "classfile/javaClasses.hpp"
28 #include "memory/allocation.inline.hpp"
29 #include "memory/filemap.hpp"
30 #include "memory/resourceArea.hpp"
31 #include "oops/oop.inline.hpp"
32 #include "runtime/safepoint.hpp"
33 #include "utilities/dtrace.hpp"
34 #include "utilities/hashtable.hpp"
35 #include "utilities/hashtable.inline.hpp"
36 #include "utilities/numberSeq.hpp"
39 // This hashtable is implemented as an open hash table with a fixed number of buckets.
41 template <MEMFLAGS F> BasicHashtableEntry<F>* BasicHashtable<F>::new_entry_free_list() {
42 BasicHashtableEntry<F>* entry = NULL;
43 if (_free_list != NULL) {
44 entry = _free_list;
45 _free_list = _free_list->next();
46 }
47 return entry;
48 }
50 // HashtableEntrys are allocated in blocks to reduce the space overhead.
51 template <MEMFLAGS F> BasicHashtableEntry<F>* BasicHashtable<F>::new_entry(unsigned int hashValue) {
52 BasicHashtableEntry<F>* entry = new_entry_free_list();
54 if (entry == NULL) {
55 if (_first_free_entry + _entry_size >= _end_block) {
56 int block_size = MIN2(512, MAX2((int)_table_size / 2, (int)_number_of_entries));
57 int len = _entry_size * block_size;
58 len = 1 << log2_intptr(len); // round down to power of 2
59 assert(len >= _entry_size, "");
60 _first_free_entry = NEW_C_HEAP_ARRAY2(char, len, F, CURRENT_PC);
61 _end_block = _first_free_entry + len;
62 }
63 entry = (BasicHashtableEntry<F>*)_first_free_entry;
64 _first_free_entry += _entry_size;
65 }
67 assert(_entry_size % HeapWordSize == 0, "");
68 entry->set_hash(hashValue);
69 return entry;
70 }
73 template <class T, MEMFLAGS F> HashtableEntry<T, F>* Hashtable<T, F>::new_entry(unsigned int hashValue, T obj) {
74 HashtableEntry<T, F>* entry;
76 entry = (HashtableEntry<T, F>*)BasicHashtable<F>::new_entry(hashValue);
77 entry->set_literal(obj);
78 return entry;
79 }
81 // Check to see if the hashtable is unbalanced. The caller set a flag to
82 // rehash at the next safepoint. If this bucket is 60 times greater than the
83 // expected average bucket length, it's an unbalanced hashtable.
84 // This is somewhat an arbitrary heuristic but if one bucket gets to
85 // rehash_count which is currently 100, there's probably something wrong.
87 template <class T, MEMFLAGS F> bool RehashableHashtable<T, F>::check_rehash_table(int count) {
88 assert(this->table_size() != 0, "underflow");
89 if (count > (((double)this->number_of_entries()/(double)this->table_size())*rehash_multiple)) {
90 // Set a flag for the next safepoint, which should be at some guaranteed
91 // safepoint interval.
92 return true;
93 }
94 return false;
95 }
97 template <class T, MEMFLAGS F> juint RehashableHashtable<T, F>::_seed = 0;
99 // Create a new table and using alternate hash code, populate the new table
100 // with the existing elements. This can be used to change the hash code
101 // and could in the future change the size of the table.
103 template <class T, MEMFLAGS F> void RehashableHashtable<T, F>::move_to(RehashableHashtable<T, F>* new_table) {
105 // Initialize the global seed for hashing.
106 _seed = AltHashing::compute_seed();
107 assert(seed() != 0, "shouldn't be zero");
109 int saved_entry_count = this->number_of_entries();
111 // Iterate through the table and create a new entry for the new table
112 for (int i = 0; i < new_table->table_size(); ++i) {
113 for (HashtableEntry<T, F>* p = this->bucket(i); p != NULL; ) {
114 HashtableEntry<T, F>* next = p->next();
115 T string = p->literal();
116 // Use alternate hashing algorithm on the symbol in the first table
117 unsigned int hashValue = string->new_hash(seed());
118 // Get a new index relative to the new table (can also change size)
119 int index = new_table->hash_to_index(hashValue);
120 p->set_hash(hashValue);
121 // Keep the shared bit in the Hashtable entry to indicate that this entry
122 // can't be deleted. The shared bit is the LSB in the _next field so
123 // walking the hashtable past these entries requires
124 // BasicHashtableEntry::make_ptr() call.
125 bool keep_shared = p->is_shared();
126 this->unlink_entry(p);
127 new_table->add_entry(index, p);
128 if (keep_shared) {
129 p->set_shared();
130 }
131 p = next;
132 }
133 }
134 // give the new table the free list as well
135 new_table->copy_freelist(this);
136 assert(new_table->number_of_entries() == saved_entry_count, "lost entry on dictionary copy?");
138 // Destroy memory used by the buckets in the hashtable. The memory
139 // for the elements has been used in a new table and is not
140 // destroyed. The memory reuse will benefit resizing the SystemDictionary
141 // to avoid a memory allocation spike at safepoint.
142 BasicHashtable<F>::free_buckets();
143 }
145 template <MEMFLAGS F> void BasicHashtable<F>::free_buckets() {
146 if (NULL != _buckets) {
147 // Don't delete the buckets in the shared space. They aren't
148 // allocated by os::malloc
149 if (!UseSharedSpaces ||
150 !FileMapInfo::current_info()->is_in_shared_space(_buckets)) {
151 FREE_C_HEAP_ARRAY(HashtableBucket, _buckets, F);
152 }
153 _buckets = NULL;
154 }
155 }
158 // Reverse the order of elements in the hash buckets.
160 template <MEMFLAGS F> void BasicHashtable<F>::reverse() {
162 for (int i = 0; i < _table_size; ++i) {
163 BasicHashtableEntry<F>* new_list = NULL;
164 BasicHashtableEntry<F>* p = bucket(i);
165 while (p != NULL) {
166 BasicHashtableEntry<F>* next = p->next();
167 p->set_next(new_list);
168 new_list = p;
169 p = next;
170 }
171 *bucket_addr(i) = new_list;
172 }
173 }
176 // Copy the table to the shared space.
178 template <MEMFLAGS F> void BasicHashtable<F>::copy_table(char** top, char* end) {
180 // Dump the hash table entries.
182 intptr_t *plen = (intptr_t*)(*top);
183 *top += sizeof(*plen);
185 int i;
186 for (i = 0; i < _table_size; ++i) {
187 for (BasicHashtableEntry<F>** p = _buckets[i].entry_addr();
188 *p != NULL;
189 p = (*p)->next_addr()) {
190 if (*top + entry_size() > end) {
191 report_out_of_shared_space(SharedMiscData);
192 }
193 *p = (BasicHashtableEntry<F>*)memcpy(*top, *p, entry_size());
194 *top += entry_size();
195 }
196 }
197 *plen = (char*)(*top) - (char*)plen - sizeof(*plen);
199 // Set the shared bit.
201 for (i = 0; i < _table_size; ++i) {
202 for (BasicHashtableEntry<F>* p = bucket(i); p != NULL; p = p->next()) {
203 p->set_shared();
204 }
205 }
206 }
210 // Reverse the order of elements in the hash buckets.
212 template <class T, MEMFLAGS F> void Hashtable<T, F>::reverse(void* boundary) {
214 for (int i = 0; i < this->table_size(); ++i) {
215 HashtableEntry<T, F>* high_list = NULL;
216 HashtableEntry<T, F>* low_list = NULL;
217 HashtableEntry<T, F>* last_low_entry = NULL;
218 HashtableEntry<T, F>* p = bucket(i);
219 while (p != NULL) {
220 HashtableEntry<T, F>* next = p->next();
221 if ((void*)p->literal() >= boundary) {
222 p->set_next(high_list);
223 high_list = p;
224 } else {
225 p->set_next(low_list);
226 low_list = p;
227 if (last_low_entry == NULL) {
228 last_low_entry = p;
229 }
230 }
231 p = next;
232 }
233 if (low_list != NULL) {
234 *bucket_addr(i) = low_list;
235 last_low_entry->set_next(high_list);
236 } else {
237 *bucket_addr(i) = high_list;
238 }
239 }
240 }
242 template <class T, MEMFLAGS F> int RehashableHashtable<T, F>::literal_size(Symbol *symbol) {
243 return symbol->size() * HeapWordSize;
244 }
246 template <class T, MEMFLAGS F> int RehashableHashtable<T, F>::literal_size(oop oop) {
247 // NOTE: this would over-count if (pre-JDK8) java_lang_Class::has_offset_field() is true,
248 // and the String.value array is shared by several Strings. However, starting from JDK8,
249 // the String.value array is not shared anymore.
250 assert(oop != NULL && oop->klass() == SystemDictionary::String_klass(), "only strings are supported");
251 return (oop->size() + java_lang_String::value(oop)->size()) * HeapWordSize;
252 }
254 // Dump footprint and bucket length statistics
255 //
256 // Note: if you create a new subclass of Hashtable<MyNewType, F>, you will need to
257 // add a new function Hashtable<T, F>::literal_size(MyNewType lit)
259 template <class T, MEMFLAGS F> void RehashableHashtable<T, F>::dump_table(outputStream* st, const char *table_name) {
260 NumberSeq summary;
261 int literal_bytes = 0;
262 for (int i = 0; i < this->table_size(); ++i) {
263 int count = 0;
264 for (HashtableEntry<T, F>* e = this->bucket(i);
265 e != NULL; e = e->next()) {
266 count++;
267 literal_bytes += literal_size(e->literal());
268 }
269 summary.add((double)count);
270 }
271 double num_buckets = summary.num();
272 double num_entries = summary.sum();
274 int bucket_bytes = (int)num_buckets * sizeof(HashtableBucket<F>);
275 int entry_bytes = (int)num_entries * sizeof(HashtableEntry<T, F>);
276 int total_bytes = literal_bytes + bucket_bytes + entry_bytes;
278 double bucket_avg = (num_buckets <= 0) ? 0 : (bucket_bytes / num_buckets);
279 double entry_avg = (num_entries <= 0) ? 0 : (entry_bytes / num_entries);
280 double literal_avg = (num_entries <= 0) ? 0 : (literal_bytes / num_entries);
282 st->print_cr("%s statistics:", table_name);
283 st->print_cr("Number of buckets : %9d = %9d bytes, avg %7.3f", (int)num_buckets, bucket_bytes, bucket_avg);
284 st->print_cr("Number of entries : %9d = %9d bytes, avg %7.3f", (int)num_entries, entry_bytes, entry_avg);
285 st->print_cr("Number of literals : %9d = %9d bytes, avg %7.3f", (int)num_entries, literal_bytes, literal_avg);
286 st->print_cr("Total footprint : %9s = %9d bytes", "", total_bytes);
287 st->print_cr("Average bucket size : %9.3f", summary.avg());
288 st->print_cr("Variance of bucket size : %9.3f", summary.variance());
289 st->print_cr("Std. dev. of bucket size: %9.3f", summary.sd());
290 st->print_cr("Maximum bucket size : %9d", (int)summary.maximum());
291 }
294 // Dump the hash table buckets.
296 template <MEMFLAGS F> void BasicHashtable<F>::copy_buckets(char** top, char* end) {
297 intptr_t len = _table_size * sizeof(HashtableBucket<F>);
298 *(intptr_t*)(*top) = len;
299 *top += sizeof(intptr_t);
301 *(intptr_t*)(*top) = _number_of_entries;
302 *top += sizeof(intptr_t);
304 if (*top + len > end) {
305 report_out_of_shared_space(SharedMiscData);
306 }
307 _buckets = (HashtableBucket<F>*)memcpy(*top, _buckets, len);
308 *top += len;
309 }
312 #ifndef PRODUCT
314 template <class T, MEMFLAGS F> void Hashtable<T, F>::print() {
315 ResourceMark rm;
317 for (int i = 0; i < BasicHashtable<F>::table_size(); i++) {
318 HashtableEntry<T, F>* entry = bucket(i);
319 while(entry != NULL) {
320 tty->print("%d : ", i);
321 entry->literal()->print();
322 tty->cr();
323 entry = entry->next();
324 }
325 }
326 }
329 template <MEMFLAGS F> void BasicHashtable<F>::verify() {
330 int count = 0;
331 for (int i = 0; i < table_size(); i++) {
332 for (BasicHashtableEntry<F>* p = bucket(i); p != NULL; p = p->next()) {
333 ++count;
334 }
335 }
336 assert(count == number_of_entries(), "number of hashtable entries incorrect");
337 }
340 #endif // PRODUCT
343 #ifdef ASSERT
345 template <MEMFLAGS F> void BasicHashtable<F>::verify_lookup_length(double load) {
346 if ((double)_lookup_length / (double)_lookup_count > load * 2.0) {
347 warning("Performance bug: SystemDictionary lookup_count=%d "
348 "lookup_length=%d average=%lf load=%f",
349 _lookup_count, _lookup_length,
350 (double) _lookup_length / _lookup_count, load);
351 }
352 }
354 #endif
355 // Explicitly instantiate these types
356 template class Hashtable<ConstantPool*, mtClass>;
357 template class RehashableHashtable<Symbol*, mtSymbol>;
358 template class RehashableHashtable<oopDesc*, mtSymbol>;
359 template class Hashtable<Symbol*, mtSymbol>;
360 template class Hashtable<Klass*, mtClass>;
361 template class Hashtable<oop, mtClass>;
362 #if defined(SOLARIS) || defined(CHECK_UNHANDLED_OOPS)
363 template class Hashtable<oop, mtSymbol>;
364 template class RehashableHashtable<oop, mtSymbol>;
365 #endif // SOLARIS || CHECK_UNHANDLED_OOPS
366 template class Hashtable<oopDesc*, mtSymbol>;
367 template class Hashtable<Symbol*, mtClass>;
368 template class HashtableEntry<Symbol*, mtSymbol>;
369 template class HashtableEntry<Symbol*, mtClass>;
370 template class HashtableEntry<oop, mtSymbol>;
371 template class BasicHashtableEntry<mtSymbol>;
372 template class BasicHashtableEntry<mtCode>;
373 template class BasicHashtable<mtClass>;
374 template class BasicHashtable<mtSymbol>;
375 template class BasicHashtable<mtCode>;
376 template class BasicHashtable<mtInternal>;