8059510: Compact symbol table layout inside shared archive

Use separate compact table for shared symbols.

Reviewed-by: iklam, gziemski, shade, sla, jrose

hotspot/src/share/vm/classfile/compactHashtable.cpp

@@ -0,0 +1,417 @@
+/*
+ * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "classfile/javaClasses.hpp"
+#include "memory/metaspaceShared.hpp"
+#include "utilities/numberSeq.hpp"
+#include <sys/stat.h>
+
+/////////////////////////////////////////////////////
+//
+// The compact hash table writer implementations
+//
+CompactHashtableWriter::CompactHashtableWriter(const char* table_name,
+                                               int num_entries,
+                                               CompactHashtableStats* stats) {
+  assert(DumpSharedSpaces, "dump-time only");
+  _table_name = table_name;
+  _num_entries = num_entries;
+  _num_buckets = number_of_buckets(_num_entries);
+  _buckets = NEW_C_HEAP_ARRAY(Entry*, _num_buckets, mtSymbol);
+  memset(_buckets, 0, sizeof(Entry*) * _num_buckets);
+
+  /* bucket sizes table */
+  _bucket_sizes = NEW_C_HEAP_ARRAY(juint, _num_buckets, mtSymbol);
+  memset(_bucket_sizes, 0, sizeof(juint) * _num_buckets);
+
+  stats->hashentry_count = _num_entries;
+  // Compact buckets' entries will have only the 4-byte offset, but
+  // we don't know how many there will be at this point. So use a
+  // conservative estimate here. The size is adjusted later when we
+  // write out the buckets.
+  stats->hashentry_bytes = _num_entries * 8;
+  stats->bucket_count    = _num_buckets;
+  stats->bucket_bytes    = (_num_buckets + 1) * (sizeof(juint));
+  _stats = stats;
+
+  // See compactHashtable.hpp for table layout
+  _required_bytes = sizeof(juint) * 2; // _base_address, written as 2 juints
+  _required_bytes+= sizeof(juint) +    // num_entries
+                    sizeof(juint) +    // num_buckets
+                    stats->hashentry_bytes +
+                    stats->bucket_bytes;
+}
+
+CompactHashtableWriter::~CompactHashtableWriter() {
+  for (int index = 0; index < _num_buckets; index++) {
+    Entry* next = NULL;
+    for (Entry* tent = _buckets[index]; tent; tent = next) {
+      next = tent->next();
+      delete tent;
+    }
+  }
+
+  FREE_C_HEAP_ARRAY(juint, _bucket_sizes);
+  FREE_C_HEAP_ARRAY(Entry*, _buckets);
+}
+
+// Calculate the number of buckets in the temporary hash table
+int CompactHashtableWriter::number_of_buckets(int num_entries) {
+  const int buksize = (int)SharedSymbolTableBucketSize;
+  int num_buckets = (num_entries + buksize - 1) / buksize;
+  num_buckets = (num_buckets + 1) & (~0x01);
+
+  return num_buckets;
+}
+
+// Add a symbol entry to the temporary hash table
+void CompactHashtableWriter::add(unsigned int hash, Entry* entry) {
+  int index = hash % _num_buckets;
+  entry->set_next(_buckets[index]);
+  _buckets[index] = entry;
+  _bucket_sizes[index] ++;
+}
+
+// Write the compact table's bucket infos
+juint* CompactHashtableWriter::dump_table(juint* p, juint** first_bucket,
+                                          NumberSeq* summary) {
+  int index;
+  juint* compact_table = p;
+  // Find the start of the buckets, skip the compact_bucket_infos table
+  // and the table end offset.
+  juint offset = _num_buckets + 1;
+  *first_bucket = compact_table + offset;
+
+  for (index = 0; index < _num_buckets; index++) {
+    int bucket_size = _bucket_sizes[index];
+    if (bucket_size == 1) {
+      // bucket with one entry is compacted and only has the symbol offset
+      compact_table[index] = BUCKET_INFO(offset, COMPACT_BUCKET_TYPE);
+      offset += bucket_size; // each entry contains symbol offset only
+    } else {
+      // regular bucket, each entry is a symbol (hash, offset) pair
+      compact_table[index] = BUCKET_INFO(offset, REGULAR_BUCKET_TYPE);
+      offset += bucket_size * 2; // each hash entry is 2 juints
+    }
+    if (offset & ~BUCKET_OFFSET_MASK) {
+      vm_exit_during_initialization("CompactHashtableWriter::dump_table: Overflow! "
+                                    "Too many symbols.");
+    }
+    summary->add(bucket_size);
+  }
+  // Mark the end of the table
+  compact_table[_num_buckets] = BUCKET_INFO(offset, TABLEEND_BUCKET_TYPE);
+
+  return compact_table;
+}
+
+// Write the compact table's entries
+juint* CompactHashtableWriter::dump_buckets(juint* compact_table, juint* p,
+                                            NumberSeq* summary) {
+  uintx base_address = uintx(MetaspaceShared::shared_rs()->base());
+  uintx max_delta    = uintx(MetaspaceShared::shared_rs()->size());
+  assert(max_delta <= 0x7fffffff, "range check");
+  int num_compact_buckets = 0;
+
+  assert(p != NULL, "sanity");
+  for (int index = 0; index < _num_buckets; index++) {
+    juint count = 0;
+    int bucket_size = _bucket_sizes[index];
+    int bucket_type = BUCKET_TYPE(compact_table[index]);
+
+    if (bucket_size == 1) {
+      assert(bucket_type == COMPACT_BUCKET_TYPE, "Bad bucket type");
+      num_compact_buckets ++;
+    }
+    for (Entry* tent = _buckets[index]; tent;
+         tent = tent->next()) {
+      if (bucket_type == REGULAR_BUCKET_TYPE) {
+        *p++ = juint(tent->hash()); // write symbol hash
+      }
+      uintx deltax = uintx(tent->value()) - base_address;
+      assert(deltax < max_delta, "range check");
+      juint delta = juint(deltax);
+      *p++ = delta; // write symbol offset
+      count ++;
+    }
+    assert(count == _bucket_sizes[index], "sanity");
+  }
+
+  // Adjust the hashentry_bytes in CompactHashtableStats. Each compact
+  // bucket saves 4-byte.
+  _stats->hashentry_bytes -= num_compact_buckets * 4;
+
+  return p;
+}
+
+// Write the compact table
+void CompactHashtableWriter::dump(char** top, char* end) {
+  NumberSeq summary;
+  char* old_top = *top;
+  juint* p = (juint*)(*top);
+
+  uintx base_address = uintx(MetaspaceShared::shared_rs()->base());
+
+  *p++ = high(base_address);
+  *p++ = low (base_address); // base address
+  *p++ = _num_entries;  // number of entries in the table
+  *p++ = _num_buckets;  // number of buckets in the table
+
+  juint* first_bucket = NULL;
+  juint* compact_table = dump_table(p, &first_bucket, &summary);
+  juint* bucket_end = dump_buckets(compact_table, first_bucket, &summary);
+
+  assert(bucket_end <= (juint*)end, "cannot write past end");
+  *top = (char*)bucket_end;
+
+  if (PrintSharedSpaces) {
+    double avg_cost = 0.0;
+    if (_num_entries > 0) {
+      avg_cost = double(_required_bytes)/double(_num_entries);
+    }
+    tty->print_cr("Shared %s table stats -------- base: " PTR_FORMAT, _table_name, (intptr_t)base_address);
+    tty->print_cr("Number of entries       : %9d", _num_entries);
+    tty->print_cr("Total bytes used        : %9d", (int)((*top) - old_top));
+    tty->print_cr("Average bytes per entry : %9.3f", avg_cost);
+    tty->print_cr("Average bucket size     : %9.3f", summary.avg());
+    tty->print_cr("Variance of bucket size : %9.3f", summary.variance());
+    tty->print_cr("Std. dev. of bucket size: %9.3f", summary.sd());
+    tty->print_cr("Maximum bucket size     : %9d", (int)summary.maximum());
+  }
+}
+
+/////////////////////////////////////////////////////////////
+//
+// The CompactHashtable implementation
+//
+template <class T, class N> const char* CompactHashtable<T, N>::init(const char* buffer) {
+  assert(!DumpSharedSpaces, "run-time only");
+  juint*p = (juint*)buffer;
+  juint upper = *p++;
+  juint lower = *p++;
+  _base_address = uintx(jlong_from(upper, lower));
+  _entry_count = *p++;
+  _bucket_count = *p++;
+  _buckets = p;
+  _table_end_offset = BUCKET_OFFSET(p[_bucket_count]); // located at the end of the bucket_info table
+
+  juint *end = _buckets + _table_end_offset;
+  return (const char*)end;
+}
+
+// Explicitly instantiate these types
+template class CompactHashtable<Symbol*, char>;
+
+#ifndef O_BINARY       // if defined (Win32) use binary files.
+#define O_BINARY 0     // otherwise do nothing.
+#endif
+
+////////////////////////////////////////////////////////
+//
+// HashtableTextDump
+//
+HashtableTextDump::HashtableTextDump(const char* filename) : _fd(-1) {
+  struct stat st;
+  if (os::stat(filename, &st) != 0) {
+    quit("Unable to get hashtable dump file size", filename);
+  }
+  _size = st.st_size;
+  _fd = open(filename, O_RDONLY | O_BINARY, 0);
+  if (_fd < 0) {
+    quit("Unable to open hashtable dump file", filename);
+  }
+  _base = os::map_memory(_fd, filename, 0, NULL, _size, true, false);
+  if (_base == NULL) {
+    quit("Unable to map hashtable dump file", filename);
+  }
+  _p = _base;
+  _end = _base + st.st_size;
+  _filename = filename;
+}
+
+HashtableTextDump::~HashtableTextDump() {
+  os::unmap_memory((char*)_base, _size);
+  if (_fd >= 0) {
+    close(_fd);
+  }
+}
+
+void HashtableTextDump::quit(const char* err, const char* msg) {
+  vm_exit_during_initialization(err, msg);
+}
+
+void HashtableTextDump::corrupted(const char *p) {
+  char info[60];
+  sprintf(info, "corrupted at pos %d", (int)(p - _base));
+  quit(info, _filename);
+}
+
+bool HashtableTextDump::skip_newline() {
+  if (_p[0] == '\r' && _p[1] == '\n') {
+    _p += 2;
+  } else if (_p[0] == '\n') {
+    _p += 1;
+  } else {
+    corrupted(_p);
+  }
+  return true;
+}
+
+int HashtableTextDump::skip(char must_be_char) {
+  corrupted_if(remain() < 1);
+  corrupted_if(*_p++ != must_be_char);
+  return 0;
+}
+
+void HashtableTextDump::skip_past(char c) {
+  for (;;) {
+    corrupted_if(remain() < 1);
+    if (*_p++ == c) {
+      return;
+    }
+  }
+}
+
+void HashtableTextDump::check_version(const char* ver) {
+  int len = (int)strlen(ver);
+  corrupted_if(remain() < len);
+  if (strncmp(_p, ver, len) != 0) {
+    quit("wrong version of hashtable dump file", _filename);
+  }
+  _p += len;
+  skip_newline();
+}
+
+
+int HashtableTextDump::scan_prefix() {
+  // Expect /[0-9]+: /
+  int utf8_length = get_num(':');
+  if (*_p != ' ') {
+    corrupted(_p);
+  }
+  _p++;
+  return utf8_length;
+}
+
+int HashtableTextDump::scan_prefix2() {
+  // Expect /[0-9]+ (-|)[0-9]+: /
+  int utf8_length = get_num(' ');
+  if (*_p == '-') {
+    _p++;
+  }
+  (void)get_num(':');
+  if (*_p != ' ') {
+    corrupted(_p);
+  }
+  _p++;
+  return utf8_length;
+}
+
+jchar HashtableTextDump::unescape(const char* from, const char* end, int count) {
+  jchar value = 0;
+
+  corrupted_if(from + count > end);
+
+  for (int i=0; i<count; i++) {
+    char c = *from++;
+    switch (c) {
+    case '0': case '1': case '2': case '3': case '4':
+    case '5': case '6': case '7': case '8': case '9':
+      value = (value << 4) + c - '0';
+      break;
+    case 'a': case 'b': case 'c':
+    case 'd': case 'e': case 'f':
+      value = (value << 4) + 10 + c - 'a';
+      break;
+    case 'A': case 'B': case 'C':
+    case 'D': case 'E': case 'F':
+      value = (value << 4) + 10 + c - 'A';
+      break;
+    default:
+      ShouldNotReachHere();
+    }
+  }
+  return value;
+}
+
+void HashtableTextDump::get_utf8(char* utf8_buffer, int utf8_length) {
+  // cache in local vars
+  const char* from = _p;
+  const char* end = _end;
+  char* to = utf8_buffer;
+  int n = utf8_length;
+
+  for (; n > 0 && from < end; n--) {
+    if (*from != '\\') {
+      *to++ = *from++;
+    } else {
+      corrupted_if(from + 2 > end);
+      char c = from[1];
+      from += 2;
+      switch (c) {
+      case 'x':
+        {
+          jchar value = unescape(from, end, 2);
+          from += 2;
+          assert(value <= 0xff, "sanity");
+          *to++ = (char)(value & 0xff);
+        }
+        break;
+      case 't':  *to++ = '\t'; break;
+      case 'n':  *to++ = '\n'; break;
+      case 'r':  *to++ = '\r'; break;
+      case '\\': *to++ = '\\'; break;
+      default:
+        ShouldNotReachHere();
+      }
+    }
+  }
+  corrupted_if(n > 0); // expected more chars but file has ended
+  _p = from;
+  skip_newline();
+}
+
+// NOTE: the content is NOT the same as
+// UTF8::as_quoted_ascii(const char* utf8_str, int utf8_length, char* buf, int buflen).
+// We want to escape \r\n\t so that output [1] is more readable; [2] can be more easily
+// parsed by scripts; [3] quickly processed by HashtableTextDump::get_utf8()
+void HashtableTextDump::put_utf8(outputStream* st, const char* utf8_string, int utf8_length) {
+  const char *c = utf8_string;
+  const char *end = c + utf8_length;
+  for (; c < end; c++) {
+    switch (*c) {
+    case '\t': st->print("\\t"); break;
+    case '\r': st->print("\\r"); break;
+    case '\n': st->print("\\n"); break;
+    case '\\': st->print("\\\\"); break;
+    default:
+      if (isprint(*c)) {
+        st->print("%c", *c);
+      } else {
+        st->print("\\x%02x", ((unsigned int)*c) & 0xff);
+      }
+    }
+  }
+}