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) 1999-2015 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 3 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, see <http://www.gnu.org/licenses/>.  */
  21
  22 #ifndef BCACHE_H
  23 #define BCACHE_H 1
  24
  25 /* A bcache is a data structure for factoring out duplication in
  26    read-only structures.  You give the bcache some string of bytes S.
  27    If the bcache already contains a copy of S, it hands you back a
  28    pointer to its copy.  Otherwise, it makes a fresh copy of S, and
  29    hands you back a pointer to that.  In either case, you can throw
  30    away your copy of S, and use the bcache's.
  31
  32    The "strings" in question are arbitrary strings of bytes --- they
  33    can contain zero bytes.  You pass in the length explicitly when you
  34    call the bcache function.
  35
  36    This means that you can put ordinary C objects in a bcache.
  37    However, if you do this, remember that structs can contain `holes'
  38    between members, added for alignment.  These bytes usually contain
  39    garbage.  If you try to bcache two objects which are identical from
  40    your code's point of view, but have different garbage values in the
  41    structure's holes, then the bcache will treat them as separate
  42    strings, and you won't get the nice elimination of duplicates you
  43    were hoping for.  So, remember to memset your structures full of
  44    zeros before bcaching them!
  45
  46    You shouldn't modify the strings you get from a bcache, because:
  47
  48    - You don't necessarily know who you're sharing space with.  If I
  49    stick eight bytes of text in a bcache, and then stick an eight-byte
  50    structure in the same bcache, there's no guarantee those two
  51    objects don't actually comprise the same sequence of bytes.  If
  52    they happen to, the bcache will use a single byte string for both
  53    of them.  Then, modifying the structure will change the string.  In
  54    bizarre ways.
  55
  56    - Even if you know for some other reason that all that's okay,
  57    there's another problem.  A bcache stores all its strings in a hash
  58    table.  If you modify a string's contents, you will probably change
  59    its hash value.  This means that the modified string is now in the
  60    wrong place in the hash table, and future bcache probes will never
  61    find it.  So by mutating a string, you give up any chance of
  62    sharing its space with future duplicates.
  63
  64
  65    Size of bcache VS hashtab:
  66
  67    For bcache, the most critical cost is size (or more exactly the
  68    overhead added by the bcache).  It turns out that the bcache is
  69    remarkably efficient.
  70
  71    Assuming a 32-bit system (the hash table slots are 4 bytes),
  72    ignoring alignment, and limit strings to 255 bytes (1 byte length)
  73    we get ...
  74
  75    bcache: This uses a separate linked list to track the hash chain.
  76    The numbers show roughly 100% occupancy of the hash table and an
  77    average chain length of 4.  Spreading the slot cost over the 4
  78    chain elements:
  79
  80    4 (slot) / 4 (chain length) + 1 (length) + 4 (chain) = 6 bytes
  81
  82    hashtab: This uses a more traditional re-hash algorithm where the
  83    chain is maintained within the hash table.  The table occupancy is
  84    kept below 75% but we'll assume its perfect:
  85
  86    4 (slot) x 4/3 (occupancy) +  1 (length) = 6 1/3 bytes
  87
  88    So a perfect hashtab has just slightly larger than an average
  89    bcache.
  90
  91    It turns out that an average hashtab is far worse.  Two things
  92    hurt:
  93
  94    - Hashtab's occupancy is more like 50% (it ranges between 38% and
  95    75%) giving a per slot cost of 4x2 vs 4x4/3.
  96
  97    - the string structure needs to be aligned to 8 bytes which for
  98    hashtab wastes 7 bytes, while for bcache wastes only 3.
  99
 100    This gives:
 101
 102    hashtab: 4 x 2 + 1 + 7 = 16 bytes
 103
 104    bcache 4 / 4 + 1 + 4 + 3 = 9 bytes
 105
 106    The numbers of GDB debugging GDB support this.  ~40% vs ~70% overhead.
 107
 108
 109    Speed of bcache VS hashtab (the half hash hack):
 110
 111    While hashtab has a typical chain length of 1, bcache has a chain
 112    length of round 4.  This means that the bcache will require
 113    something like double the number of compares after that initial
 114    hash.  In both cases the comparison takes the form:
 115
 116    a.length == b.length && memcmp (a.data, b.data, a.length) == 0
 117
 118    That is lengths are checked before doing the memcmp.
 119
 120    For GDB debugging GDB, it turned out that all lengths were 24 bytes
 121    (no C++ so only psymbols were cached) and hence, all compares
 122    required a call to memcmp.  As a hack, two bytes of padding
 123    (mentioned above) are used to store the upper 16 bits of the
 124    string's hash value and then that is used in the comparison vis:
 125
 126    a.half_hash == b.half_hash && a.length == b.length && memcmp
 127    (a.data, b.data, a.length)
 128
 129    The numbers from GDB debugging GDB show this to be a remarkable
 130    100% effective (only necessary length and memcmp tests being
 131    performed).
 132
 133    Mind you, looking at the wall clock, the same GDB debugging GDB
 134    showed only marginal speed up (0.780 vs 0.773s).  Seems GDB is too
 135    busy doing something else :-(
 136
 137 */
 138
 139
 140 struct bcache;
 141
 142 /* Find a copy of the LENGTH bytes at ADDR in BCACHE.  If BCACHE has
 143    never seen those bytes before, add a copy of them to BCACHE.  In
 144    either case, return a pointer to BCACHE's copy of that string.
 145    Since the cached value is ment to be read-only, return a const
 146    buffer.  */
 147 extern const void *bcache (const void *addr, int length,
 148                            struct bcache *bcache);
 149
 150 /* Like bcache, but if ADDED is not NULL, set *ADDED to true if the
 151    bytes were newly added to the cache, or to false if the bytes were
 152    found in the cache.  */
 153 extern const void *bcache_full (const void *addr, int length,
 154                                 struct bcache *bcache, int *added);
 155
 156 /* Free all the storage used by BCACHE.  */
 157 extern void bcache_xfree (struct bcache *bcache);
 158
 159 /* Create a new bcache object.  */
 160 extern struct bcache *bcache_xmalloc (
 161     unsigned long (*hash_function)(const void *, int length),
 162     int (*compare_function)(const void *, const void *, int length));
 163
 164 /* Print statistics on BCACHE's memory usage and efficacity at
 165    eliminating duplication.  TYPE should be a string describing the
 166    kind of data BCACHE holds.  Statistics are printed using
 167    `printf_filtered' and its ilk.  */
 168 extern void print_bcache_statistics (struct bcache *bcache, char *type);
 169 extern int bcache_memory_used (struct bcache *bcache);
 170
 171 /* The hash functions */
 172 extern unsigned long hash(const void *addr, int length);
 173 extern unsigned long hash_continue (const void *addr, int length,
 174                                     unsigned long h);
 175
 176 #endif /* BCACHE_H */