[Zlib-devel] Some inflate_fast() statistics

[Mark Adler]

Steve and other interested nerds,

Here are some match length frequency counts from typical deflate  
data, generated using the infgen and pngdat utilities that I sent on  
this email list a while back, and standard Unix utilities, thusly:

     infgen < test.tar.gz | grep ^match | awk '{ print $2 }' | sort - 
n | uniq -c

or:

     pngdat < bigpngs/pdp11,70.png | infgen | grep ^match | awk  
'{ print $2 }' | sort -n | uniq -c

mark


"Typical" web pages stored in a tar file (the large number of 258  
matches is explained by the series of nulls in tar files):

gromit% infgen < test.tar.gz | grep ^match | awk '{ print $2 }' |  
sort -n | uniq -c
70632 3
96627 4
66200 5
50093 6
36953 7
29365 8
22928 9
20179 10
16400 11
13893 12
11261 13
8429 14
7812 15
5977 16
5508 17
4923 18
4209 19
3407 20
3562 21
3335 22
2909 23
2559 24
2589 25
2498 26
2275 27
2538 28
2101 29
2238 30
1547 31
1819 32
1258 33
1446 34
1407 35
1115 36
1017 37
1093 38
1046 39
1223 40
1215 41
1282 42
1274 43
977 44
1061 45
978 46
763 47
623 48
635 49
753 50
502 51
656 52
499 53
618 54
471 55
416 56
417 57
414 58
812 59
325 60
333 61
452 62
588 63
627 64
460 65
240 66
319 67
246 68
196 69
282 70
198 71
196 72
306 73
375 74
127 75
199 76
154 77
231 78
173 79
194 80
218 81
274 82
188 83
218 84
159 85
160 86
254 87
188 88
151 89
354 90
178 91
198 92
109 93
100 94
127 95
133 96
139 97
   88 98
112 99
179 100
202 101
105 102
113 103
   98 104
107 105
141 106
159 107
108 108
   93 109
   81 110
   48 111
   42 112
150 113
346 114
183 115
177 116
   96 117
   41 118
   87 119
   41 120
   67 121
104 122
   71 123
   97 124
   50 125
   90 126
   76 127
   61 128
   67 129
   63 130
   85 131
   50 132
140 133
102 134
   89 135
   89 136
   76 137
   59 138
   70 139
   92 140
   76 141
   60 142
   50 143
   63 144
   58 145
   37 146
   48 147
   41 148
   72 149
   67 150
   86 151
   70 152
   80 153
   51 154
   77 155
118 156
   39 157
   43 158
   42 159
   49 160
   17 161
   25 162
   69 163
   26 164
   52 165
   44 166
   52 167
   61 168
   68 169
   37 170
   28 171
   64 172
   29 173
   41 174
   28 175
   34 176
   45 177
   31 178
   41 179
   38 180
   67 181
   35 182
   23 183
   22 184
   25 185
   35 186
   21 187
   34 188
   24 189
   27 190
   37 191
   58 192
   31 193
   26 194
   49 195
   26 196
   25 197
   17 198
   30 199
   31 200
   17 201
   28 202
    6 203
   21 204
   12 205
   12 206
   17 207
   25 208
   41 209
   77 210
   19 211
    8 212
   11 213
   14 214
   20 215
   26 216
   18 217
   26 218
   19 219
   22 220
   15 221
   14 222
   15 223
   21 224
   12 225
   10 226
   27 227
   25 228
   32 229
   11 230
   19 231
   16 232
   32 233
   39 234
   38 235
    7 236
   10 237
   19 238
   16 239
   15 240
   13 241
   22 242
   39 243
    9 244
    8 245
   14 246
    9 247
    7 248
   21 249
   16 250
    6 251
   22 252
   10 253
    9 254
   27 255
   12 256
   51 257
3367 258

A large png image (Steve's favorite):

gromit% pngdat < bigpngs/pdp11,70.png | infgen | grep ^match | awk  
'{ print $2 }' | sort -n | uniq -c
589718 3
1225451 4
705512 5
355213 6
220095 7
170913 8
132174 9
102171 10
85400 11
66028 12
47291 13
36402 14
26569 15
18257 16
13741 17
10086 18
6587 19
4848 20
3663 21
2356 22
1731 23
1499 24
916 25
694 26
577 27
342 28
288 29
233 30
127 31
137 32
120 33
   71 34
   56 35
   46 36
   32 37
   33 38
   31 39
   17 40
   10 41
    9 42
   15 43
    6 44
    9 45
    2 46
    3 47
    4 48
    1 50
    4 51
    1 52
    1 53
    1 59