gdb/bcache.c

   1 /* Implement a cached obstack.
   2    Written by Fred Fish <fnf@cygnus.com>
   3    Rewritten by Jim Blandy <jimb@cygnus.com>
   4    Copyright 1999 Free Software Foundation, Inc.
   5
   6    This file is part of GDB.
   7
   8    This program is free software; you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation; either version 2 of the License, or
  11    (at your option) any later version.
  12
  13    This program is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License
  19    along with this program; if not, write to the Free Software
  20    Foundation, Inc., 59 Temple Place - Suite 330,
  21    Boston, MA 02111-1307, USA.  */
  22
  23 #include <stddef.h>
  24 #include <stdlib.h>
  25
  26 #include "defs.h"
  27 #include "obstack.h"
  28 #include "bcache.h"
  29 #include "gdb_string.h"         /* For memcpy declaration */
  30
  31
  32 \f
  33 /* The hash function.  */
  34
  35 unsigned long
  36 hash (void *addr, int length)
  37 {
  38   /* If it's a short string, hash on every character.  Otherwise, sample
  39      characters from throughout the string.  */
  40   if (length <= 64)
  41     {
  42       char *byte = addr;
  43       unsigned long h = 0;
  44       int i;
  45
  46       for (i = 0; i < length; i++)
  47         h = h * 65793 ^ (h >> (sizeof (h) * 8 - 6)) ^ byte[i];
  48
  49       return h;
  50     }
  51   else
  52     {
  53       char *byte = addr;
  54       int n, i;
  55       unsigned long h = 0;
  56
  57       for (n = i = 0; n < 64; n++)
  58         {
  59           h = h * 65793 + (h >> (sizeof (h) * 8 - 6)) + byte[i];
  60           i = h % length;
  61         }
  62
  63       return h;
  64     }
  65 }
  66
  67 \f
  68 /* Growing the bcache's hash table.  */
  69
  70 /* If the average chain length grows beyond this, then we want to
  71    resize our hash table.  */
  72 #define CHAIN_LENGTH_THRESHOLD (5)
  73
  74 static void
  75 expand_hash_table (struct bcache *bcache)
  76 {
  77   /* A table of good hash table sizes.  Whenever we grow, we pick the
  78      next larger size from this table.  sizes[i] is close to 1 << (i+10),
  79      so we roughly double the table size each time.  After we fall off
  80      the end of this table, we just double.  Don't laugh --- there have
  81      been executables sighted with a gigabyte of debug info.  */
  82   static unsigned long sizes[] = {
  83     1021, 2053, 4099, 8191, 16381, 32771,
  84     65537, 131071, 262144, 524287, 1048573, 2097143,
  85     4194301, 8388617, 16777213, 33554467, 67108859, 134217757,
  86     268435459, 536870923, 1073741827, 2147483659UL
  87   };
  88   int new_num_buckets;
  89   struct bstring **new_buckets;
  90   int i;
  91
  92   /* Find the next size.  */
  93   for (i = 0; i < (sizeof (sizes) / sizeof (sizes[0])); i++)
  94     if (sizes[i] > bcache->num_buckets)
  95       {
  96         new_num_buckets = sizes[i];
  97         break;
  98       }
  99   if (i >= (sizeof (sizes) / sizeof (sizes[0])))
 100     new_num_buckets = bcache->num_buckets * 2;
 101
 102   /* Allocate the new table.  */
 103   {
 104     size_t new_size = new_num_buckets * sizeof (new_buckets[0]);
 105     new_buckets = (struct bstring **) xmalloc (new_size);
 106     memset (new_buckets, 0, new_size);
 107
 108     bcache->structure_size -= (bcache->num_buckets
 109                                * sizeof (bcache->bucket[0]));
 110     bcache->structure_size += new_size;
 111   }
 112
 113   /* Rehash all existing strings.  */
 114   for (i = 0; i < bcache->num_buckets; i++)
 115     {
 116       struct bstring *s, *next;
 117
 118       for (s = bcache->bucket[i]; s; s = next)
 119         {
 120           struct bstring **new_bucket;
 121           next = s->next;
 122
 123           new_bucket = &new_buckets[(hash (&s->d.data, s->length)
 124                                      % new_num_buckets)];
 125           s->next = *new_bucket;
 126           *new_bucket = s;
 127         }
 128     }
 129
 130   /* Plug in the new table.  */
 131   if (bcache->bucket)
 132     free (bcache->bucket);
 133   bcache->bucket = new_buckets;
 134   bcache->num_buckets = new_num_buckets;
 135 }
 136
 137 \f
 138 /* Looking up things in the bcache.  */
 139
 140 /* The number of bytes needed to allocate a struct bstring whose data
 141    is N bytes long.  */
 142 #define BSTRING_SIZE(n) (offsetof (struct bstring, d.data) + (n))
 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 void *
 148 bcache (void *addr, int length, struct bcache *bcache)
 149 {
 150   int hash_index;
 151   struct bstring *s;
 152
 153   /* If our average chain length is too high, expand the hash table.  */
 154   if (bcache->unique_count >= bcache->num_buckets * CHAIN_LENGTH_THRESHOLD)
 155     expand_hash_table (bcache);
 156
 157   bcache->total_count++;
 158   bcache->total_size += length;
 159
 160   hash_index = hash (addr, length) % bcache->num_buckets;
 161
 162   /* Search the hash bucket for a string identical to the caller's.  */
 163   for (s = bcache->bucket[hash_index]; s; s = s->next)
 164     if (s->length == length
 165         && ! memcmp (&s->d.data, addr, length))
 166       return &s->d.data;
 167
 168   /* The user's string isn't in the list.  Insert it after *ps.  */
 169   {
 170     struct bstring *new
 171       = obstack_alloc (&bcache->cache, BSTRING_SIZE (length));
 172     memcpy (&new->d.data, addr, length);
 173     new->length = length;
 174     new->next = bcache->bucket[hash_index];
 175     bcache->bucket[hash_index] = new;
 176
 177     bcache->unique_count++;
 178     bcache->unique_size += length;
 179     bcache->structure_size += BSTRING_SIZE (length);
 180
 181     return &new->d.data;
 182   }
 183 }
 184
 185 \f
 186 /* Freeing bcaches.  */
 187
 188 /* Free all the storage associated with BCACHE.  */
 189 void
 190 free_bcache (struct bcache *bcache)
 191 {
 192   obstack_free (&bcache->cache, 0);
 193   free (bcache->bucket);
 194
 195   /* This isn't necessary, but at least the bcache is always in a
 196      consistent state.  */
 197   memset (bcache, 0, sizeof (*bcache));
 198 }
 199
 200
 201 \f
 202 /* Printing statistics.  */
 203
 204 static int
 205 compare_ints (const void *ap, const void *bp)
 206 {
 207   /* Because we know we're comparing two ints which are positive,
 208      there's no danger of overflow here.  */
 209   return * (int *) ap - * (int *) bp;
 210 }
 211
 212
 213 static void
 214 print_percentage (int portion, int total)
 215 {
 216   if (total == 0)
 217     printf_filtered ("(not applicable)\n");
 218   else
 219     printf_filtered ("%3d%%\n", portion * 100 / total);
 220 }
 221
 222
 223 /* Print statistics on BCACHE's memory usage and efficacity at
 224    eliminating duplication.  NAME should describe the kind of data
 225    BCACHE holds.  Statistics are printed using `printf_filtered' and
 226    its ilk.  */
 227 void
 228 print_bcache_statistics (struct bcache *c, char *type)
 229 {
 230   int occupied_buckets;
 231   int max_chain_length;
 232   int median_chain_length;
 233
 234   /* Count the number of occupied buckets, and measure chain lengths.  */
 235   {
 236     int b;
 237     int *chain_length
 238       = (int *) alloca (c->num_buckets * sizeof (*chain_length));
 239
 240     occupied_buckets = 0;
 241
 242     for (b = 0; b < c->num_buckets; b++)
 243       {
 244         struct bstring *s = c->bucket[b];
 245
 246         chain_length[b] = 0;
 247
 248         if (s)
 249           {
 250             occupied_buckets++;
 251
 252             while (s)
 253               {
 254                 chain_length[b]++;
 255                 s = s->next;
 256               }
 257           }
 258       }
 259
 260     /* To compute the median, we need the set of chain lengths sorted.  */
 261     qsort (chain_length, c->num_buckets, sizeof (chain_length[0]),
 262            compare_ints);
 263
 264     if (c->num_buckets > 0)
 265       {
 266         max_chain_length = chain_length[c->num_buckets - 1];
 267         median_chain_length = chain_length[c->num_buckets / 2];
 268       }
 269     else
 270       {
 271         max_chain_length = 0;
 272         median_chain_length = 0;
 273       }
 274   }
 275
 276   printf_filtered ("  Cached '%s' statistics:\n", type);
 277   printf_filtered ("    Total object count:  %ld\n", c->total_count);
 278   printf_filtered ("    Unique object count: %ld\n", c->unique_count);
 279   printf_filtered ("    Percentage of duplicates, by count: ");
 280   print_percentage (c->total_count - c->unique_count, c->total_count);
 281   printf_filtered ("\n");
 282
 283   printf_filtered ("    Total object size:   %ld\n", c->total_size);
 284   printf_filtered ("    Unique object size:  %ld\n", c->unique_size);
 285   printf_filtered ("    Percentage of duplicates, by size:  ");
 286   print_percentage (c->total_size - c->unique_size, c->total_size);
 287   printf_filtered ("\n");
 288
 289   printf_filtered ("    Total memory used by bcache, including overhead: %ld\n",
 290                    c->structure_size);
 291   printf_filtered ("    Percentage memory overhead: ");
 292   print_percentage (c->structure_size - c->unique_size, c->unique_size);
 293   printf_filtered ("    Net memory savings:         ");
 294   print_percentage (c->total_size - c->structure_size, c->total_size);
 295   printf_filtered ("\n");
 296
 297   printf_filtered ("    Hash table size:           %3d\n", c->num_buckets);
 298   printf_filtered ("    Hash table population:     ");
 299   print_percentage (occupied_buckets, c->num_buckets);
 300   printf_filtered ("    Median hash chain length:  %3d\n",
 301                    median_chain_length);
 302   printf_filtered ("    Average hash chain length: ");
 303   if (c->num_buckets > 0)
 304     printf_filtered ("%3ld\n", c->unique_count / c->num_buckets);
 305   else
 306     printf_filtered ("(not applicable)\n");
 307   printf_filtered ("    Maximum hash chain length: %3d\n", max_chain_length);
 308   printf_filtered ("\n");
 309 }