gdb/bcache.h

   1 /* Include file cached obstack implementation.
   2    Written by Fred Fish <fnf@cygnus.com>
   3    Rewritten by Jim Blandy <jimb@cygnus.com>
   4
   5    Copyright (C) 1999, 2000, 2002, 2003 Free Software Foundation, Inc.
   6
   7    This file is part of GDB.
   8
   9    This program is free software; you can redistribute it and/or modify
  10    it under the terms of the GNU General Public License as published by
  11    the Free Software Foundation; either version 2 of the License, or
  12    (at your option) any later version.
  13
  14    This program is distributed in the hope that it will be useful,
  15    but WITHOUT ANY WARRANTY; without even the implied warranty of
  16    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17    GNU General Public License for more details.
  18
  19    You should have received a copy of the GNU General Public License
  20    along with this program; if not, write to the Free Software
  21    Foundation, Inc., 51 Franklin Street, Fifth Floor,
  22    Boston, MA 02110-1301, USA.  */
  23
  24 #ifndef BCACHE_H
  25 #define BCACHE_H 1
  26
  27 /* A bcache is a data structure for factoring out duplication in
  28    read-only structures.  You give the bcache some string of bytes S.
  29    If the bcache already contains a copy of S, it hands you back a
  30    pointer to its copy.  Otherwise, it makes a fresh copy of S, and
  31    hands you back a pointer to that.  In either case, you can throw
  32    away your copy of S, and use the bcache's.
  33
  34    The "strings" in question are arbitrary strings of bytes --- they
  35    can contain zero bytes.  You pass in the length explicitly when you
  36    call the bcache function.
  37
  38    This means that you can put ordinary C objects in a bcache.
  39    However, if you do this, remember that structs can contain `holes'
  40    between members, added for alignment.  These bytes usually contain
  41    garbage.  If you try to bcache two objects which are identical from
  42    your code's point of view, but have different garbage values in the
  43    structure's holes, then the bcache will treat them as separate
  44    strings, and you won't get the nice elimination of duplicates you
  45    were hoping for.  So, remember to memset your structures full of
  46    zeros before bcaching them!
  47
  48    You shouldn't modify the strings you get from a bcache, because:
  49
  50    - You don't necessarily know who you're sharing space with.  If I
  51    stick eight bytes of text in a bcache, and then stick an eight-byte
  52    structure in the same bcache, there's no guarantee those two
  53    objects don't actually comprise the same sequence of bytes.  If
  54    they happen to, the bcache will use a single byte string for both
  55    of them.  Then, modifying the structure will change the string.  In
  56    bizarre ways.
  57
  58    - Even if you know for some other reason that all that's okay,
  59    there's another problem.  A bcache stores all its strings in a hash
  60    table.  If you modify a string's contents, you will probably change
  61    its hash value.  This means that the modified string is now in the
  62    wrong place in the hash table, and future bcache probes will never
  63    find it.  So by mutating a string, you give up any chance of
  64    sharing its space with future duplicates.
  65
  66
  67    Size of bcache VS hashtab:
  68
  69    For bcache, the most critical cost is size (or more exactly the
  70    overhead added by the bcache).  It turns out that the bcache is
  71    remarkably efficient.
  72
  73    Assuming a 32-bit system (the hash table slots are 4 bytes),
  74    ignoring alignment, and limit strings to 255 bytes (1 byte length)
  75    we get ...
  76
  77    bcache: This uses a separate linked list to track the hash chain.
  78    The numbers show roughly 100% occupancy of the hash table and an
  79    average chain length of 4.  Spreading the slot cost over the 4
  80    chain elements:
  81
  82    4 (slot) / 4 (chain length) + 1 (length) + 4 (chain) = 6 bytes
  83
  84    hashtab: This uses a more traditional re-hash algorithm where the
  85    chain is maintained within the hash table.  The table occupancy is
  86    kept below 75% but we'll assume its perfect:
  87
  88    4 (slot) x 4/3 (occupancy) +  1 (length) = 6 1/3 bytes
  89
  90    So a perfect hashtab has just slightly larger than an average
  91    bcache.
  92
  93    It turns out that an average hashtab is far worse.  Two things
  94    hurt:
  95
  96    - Hashtab's occupancy is more like 50% (it ranges between 38% and
  97    75%) giving a per slot cost of 4x2 vs 4x4/3.
  98
  99    - the string structure needs to be aligned to 8 bytes which for
 100    hashtab wastes 7 bytes, while for bcache wastes only 3.
 101
 102    This gives:
 103
 104    hashtab: 4 x 2 + 1 + 7 = 16 bytes
 105
 106    bcache 4 / 4 + 1 + 4 + 3 = 9 bytes
 107
 108    The numbers of GDB debugging GDB support this.  ~40% vs ~70% overhead.
 109
 110
 111    Speed of bcache VS hashtab (the half hash hack):
 112
 113    While hashtab has a typical chain length of 1, bcache has a chain
 114    length of round 4.  This means that the bcache will require
 115    something like double the number of compares after that initial
 116    hash.  In both cases the comparison takes the form:
 117
 118    a.length == b.length && memcmp (a.data, b.data, a.length) == 0
 119
 120    That is lengths are checked before doing the memcmp.
 121
 122    For GDB debugging GDB, it turned out that all lengths were 24 bytes
 123    (no C++ so only psymbols were cached) and hence, all compares
 124    required a call to memcmp.  As a hack, two bytes of padding
 125    (mentioned above) are used to store the upper 16 bits of the
 126    string's hash value and then that is used in the comparison vis:
 127
 128    a.half_hash == b.half_hash && a.length == b.length && memcmp
 129    (a.data, b.data, a.length)
 130
 131    The numbers from GDB debugging GDB show this to be a remarkable
 132    100% effective (only necessary length and memcmp tests being
 133    performed).
 134
 135    Mind you, looking at the wall clock, the same GDB debugging GDB
 136    showed only marginal speed up (0.780 vs 0.773s).  Seems GDB is too
 137    busy doing something else :-(
 138
 139 */
 140
 141
 142 struct bcache;
 143
 144 /* Find a copy of the LENGTH bytes at ADDR in BCACHE.  If BCACHE has
 145    never seen those bytes before, add a copy of them to BCACHE.  In
 146    either case, return a pointer to BCACHE's copy of that string.
 147    Since the cached value is ment to be read-only, return a const
 148    buffer.  */
 149 extern void *deprecated_bcache (const void *addr, int length,
 150                                 struct bcache *bcache);
 151 extern const void *bcache (const void *addr, int length,
 152                            struct bcache *bcache);
 153
 154 /* Free all the storage used by BCACHE.  */
 155 extern void bcache_xfree (struct bcache *bcache);
 156
 157 /* Create a new bcache object.  */
 158 extern struct bcache *bcache_xmalloc (void);
 159
 160 /* Print statistics on BCACHE's memory usage and efficacity at
 161    eliminating duplication.  TYPE should be a string describing the
 162    kind of data BCACHE holds.  Statistics are printed using
 163    `printf_filtered' and its ilk.  */
 164 extern void print_bcache_statistics (struct bcache *bcache, char *type);
 165 extern int bcache_memory_used (struct bcache *bcache);
 166
 167 /* The hash function */
 168 extern unsigned long hash(const void *addr, int length);
 169
 170 #endif /* BCACHE_H */