|
7ef7284…
|
drh
|
1 |
/* enough.c -- determine the maximum size of inflate's Huffman code tables over |
|
adb9e8e…
|
drh
|
2 |
* all possible valid and complete prefix codes, subject to a length limit. |
|
6ea30fb…
|
florian
|
3 |
* Copyright (C) 2007, 2008, 2012, 2018, 2024 Mark Adler |
|
6ea30fb…
|
florian
|
4 |
* Version 1.6 29 July 2024 Mark Adler |
|
7ef7284…
|
drh
|
5 |
*/ |
|
7ef7284…
|
drh
|
6 |
|
|
7ef7284…
|
drh
|
7 |
/* Version history: |
|
7ef7284…
|
drh
|
8 |
1.0 3 Jan 2007 First version (derived from codecount.c version 1.4) |
|
7ef7284…
|
drh
|
9 |
1.1 4 Jan 2007 Use faster incremental table usage computation |
|
7ef7284…
|
drh
|
10 |
Prune examine() search on previously visited states |
|
7ef7284…
|
drh
|
11 |
1.2 5 Jan 2007 Comments clean up |
|
7ef7284…
|
drh
|
12 |
As inflate does, decrease root for short codes |
|
7ef7284…
|
drh
|
13 |
Refuse cases where inflate would increase root |
|
7ef7284…
|
drh
|
14 |
1.3 17 Feb 2008 Add argument for initial root table size |
|
7ef7284…
|
drh
|
15 |
Fix bug for initial root table size == max - 1 |
|
7ef7284…
|
drh
|
16 |
Use a macro to compute the history index |
|
bb4776e…
|
jan.nijtmans
|
17 |
1.4 18 Aug 2012 Avoid shifts more than bits in type (caused endless loop!) |
|
bb4776e…
|
jan.nijtmans
|
18 |
Clean up comparisons of different types |
|
bb4776e…
|
jan.nijtmans
|
19 |
Clean up code indentation |
|
adb9e8e…
|
drh
|
20 |
1.5 5 Aug 2018 Clean up code style, formatting, and comments |
|
adb9e8e…
|
drh
|
21 |
Show all the codes for the maximum, and only the maximum |
|
6ea30fb…
|
florian
|
22 |
1.6 29 Jul 2024 Avoid use of uintmax_t |
|
7ef7284…
|
drh
|
23 |
*/ |
|
7ef7284…
|
drh
|
24 |
|
|
7ef7284…
|
drh
|
25 |
/* |
|
adb9e8e…
|
drh
|
26 |
Examine all possible prefix codes for a given number of symbols and a |
|
adb9e8e…
|
drh
|
27 |
maximum code length in bits to determine the maximum table size for zlib's |
|
adb9e8e…
|
drh
|
28 |
inflate. Only complete prefix codes are counted. |
|
7ef7284…
|
drh
|
29 |
|
|
7ef7284…
|
drh
|
30 |
Two codes are considered distinct if the vectors of the number of codes per |
|
adb9e8e…
|
drh
|
31 |
length are not identical. So permutations of the symbol assignments result |
|
7ef7284…
|
drh
|
32 |
in the same code for the counting, as do permutations of the assignments of |
|
7ef7284…
|
drh
|
33 |
the bit values to the codes (i.e. only canonical codes are counted). |
|
7ef7284…
|
drh
|
34 |
|
|
7ef7284…
|
drh
|
35 |
We build a code from shorter to longer lengths, determining how many symbols |
|
adb9e8e…
|
drh
|
36 |
are coded at each length. At each step, we have how many symbols remain to |
|
7ef7284…
|
drh
|
37 |
be coded, what the last code length used was, and how many bit patterns of |
|
7ef7284…
|
drh
|
38 |
that length remain unused. Then we add one to the code length and double the |
|
adb9e8e…
|
drh
|
39 |
number of unused patterns to graduate to the next code length. We then |
|
7ef7284…
|
drh
|
40 |
assign all portions of the remaining symbols to that code length that |
|
adb9e8e…
|
drh
|
41 |
preserve the properties of a correct and eventually complete code. Those |
|
7ef7284…
|
drh
|
42 |
properties are: we cannot use more bit patterns than are available; and when |
|
adb9e8e…
|
drh
|
43 |
all the symbols are used, there are exactly zero possible bit patterns left |
|
adb9e8e…
|
drh
|
44 |
unused. |
|
7ef7284…
|
drh
|
45 |
|
|
7ef7284…
|
drh
|
46 |
The inflate Huffman decoding algorithm uses two-level lookup tables for |
|
adb9e8e…
|
drh
|
47 |
speed. There is a single first-level table to decode codes up to root bits |
|
adb9e8e…
|
drh
|
48 |
in length (root == 9 for literal/length codes and root == 6 for distance |
|
adb9e8e…
|
drh
|
49 |
codes, in the current inflate implementation). The base table has 1 << root |
|
adb9e8e…
|
drh
|
50 |
entries and is indexed by the next root bits of input. Codes shorter than |
|
adb9e8e…
|
drh
|
51 |
root bits have replicated table entries, so that the correct entry is |
|
adb9e8e…
|
drh
|
52 |
pointed to regardless of the bits that follow the short code. If the code is |
|
adb9e8e…
|
drh
|
53 |
longer than root bits, then the table entry points to a second-level table. |
|
adb9e8e…
|
drh
|
54 |
The size of that table is determined by the longest code with that root-bit |
|
adb9e8e…
|
drh
|
55 |
prefix. If that longest code has length len, then the table has size 1 << |
|
adb9e8e…
|
drh
|
56 |
(len - root), to index the remaining bits in that set of codes. Each |
|
adb9e8e…
|
drh
|
57 |
subsequent root-bit prefix then has its own sub-table. The total number of |
|
adb9e8e…
|
drh
|
58 |
table entries required by the code is calculated incrementally as the number |
|
adb9e8e…
|
drh
|
59 |
of codes at each bit length is populated. When all of the codes are shorter |
|
adb9e8e…
|
drh
|
60 |
than root bits, then root is reduced to the longest code length, resulting |
|
adb9e8e…
|
drh
|
61 |
in a single, smaller, one-level table. |
|
7ef7284…
|
drh
|
62 |
|
|
7ef7284…
|
drh
|
63 |
The inflate algorithm also provides for small values of root (relative to |
|
7ef7284…
|
drh
|
64 |
the log2 of the number of symbols), where the shortest code has more bits |
|
adb9e8e…
|
drh
|
65 |
than root. In that case, root is increased to the length of the shortest |
|
adb9e8e…
|
drh
|
66 |
code. This program, by design, does not handle that case, so it is verified |
|
adb9e8e…
|
drh
|
67 |
that the number of symbols is less than 1 << (root + 1). |
|
7ef7284…
|
drh
|
68 |
|
|
7ef7284…
|
drh
|
69 |
In order to speed up the examination (by about ten orders of magnitude for |
|
7ef7284…
|
drh
|
70 |
the default arguments), the intermediate states in the build-up of a code |
|
adb9e8e…
|
drh
|
71 |
are remembered and previously visited branches are pruned. The memory |
|
7ef7284…
|
drh
|
72 |
required for this will increase rapidly with the total number of symbols and |
|
adb9e8e…
|
drh
|
73 |
the maximum code length in bits. However this is a very small price to pay |
|
7ef7284…
|
drh
|
74 |
for the vast speedup. |
|
7ef7284…
|
drh
|
75 |
|
|
adb9e8e…
|
drh
|
76 |
First, all of the possible prefix codes are counted, and reachable |
|
7ef7284…
|
drh
|
77 |
intermediate states are noted by a non-zero count in a saved-results array. |
|
7ef7284…
|
drh
|
78 |
Second, the intermediate states that lead to (root + 1) bit or longer codes |
|
7ef7284…
|
drh
|
79 |
are used to look at all sub-codes from those junctures for their inflate |
|
adb9e8e…
|
drh
|
80 |
memory usage. (The amount of memory used is not affected by the number of |
|
7ef7284…
|
drh
|
81 |
codes of root bits or less in length.) Third, the visited states in the |
|
7ef7284…
|
drh
|
82 |
construction of those sub-codes and the associated calculation of the table |
|
7ef7284…
|
drh
|
83 |
size is recalled in order to avoid recalculating from the same juncture. |
|
7ef7284…
|
drh
|
84 |
Beginning the code examination at (root + 1) bit codes, which is enabled by |
|
7ef7284…
|
drh
|
85 |
identifying the reachable nodes, accounts for about six of the orders of |
|
adb9e8e…
|
drh
|
86 |
magnitude of improvement for the default arguments. About another four |
|
adb9e8e…
|
drh
|
87 |
orders of magnitude come from not revisiting previous states. Out of |
|
adb9e8e…
|
drh
|
88 |
approximately 2x10^16 possible prefix codes, only about 2x10^6 sub-codes |
|
7ef7284…
|
drh
|
89 |
need to be examined to cover all of the possible table memory usage cases |
|
7ef7284…
|
drh
|
90 |
for the default arguments of 286 symbols limited to 15-bit codes. |
|
7ef7284…
|
drh
|
91 |
|
|
6ea30fb…
|
florian
|
92 |
Note that unsigned long long is used for counting. It is quite easy to |
|
adb9e8e…
|
drh
|
93 |
exceed the capacity of an eight-byte integer with a large number of symbols |
|
adb9e8e…
|
drh
|
94 |
and a large maximum code length, so multiple-precision arithmetic would need |
|
adb9e8e…
|
drh
|
95 |
to replace the integer arithmetic in that case. This program will abort if |
|
adb9e8e…
|
drh
|
96 |
an overflow occurs. The big_t type identifies where the counting takes |
|
adb9e8e…
|
drh
|
97 |
place. |
|
adb9e8e…
|
drh
|
98 |
|
|
6ea30fb…
|
florian
|
99 |
The unsigned long long type is also used for calculating the number of |
|
6ea30fb…
|
florian
|
100 |
possible codes remaining at the maximum length. This limits the maximum code |
|
6ea30fb…
|
florian
|
101 |
length to the number of bits in a long long minus the number of bits needed |
|
6ea30fb…
|
florian
|
102 |
to represent the symbols in a flat code. The code_t type identifies where |
|
6ea30fb…
|
florian
|
103 |
the bit-pattern counting takes place. |
|
7ef7284…
|
drh
|
104 |
*/ |
|
7ef7284…
|
drh
|
105 |
|
|
7ef7284…
|
drh
|
106 |
#include <stdio.h> |
|
7ef7284…
|
drh
|
107 |
#include <stdlib.h> |
|
7ef7284…
|
drh
|
108 |
#include <string.h> |
|
adb9e8e…
|
drh
|
109 |
#include <stdarg.h> |
|
7ef7284…
|
drh
|
110 |
#include <assert.h> |
|
7ef7284…
|
drh
|
111 |
|
|
7ef7284…
|
drh
|
112 |
#define local static |
|
7ef7284…
|
drh
|
113 |
|
|
adb9e8e…
|
drh
|
114 |
// Special data types. |
|
6ea30fb…
|
florian
|
115 |
typedef unsigned long long big_t; // type for code counting |
|
6ea30fb…
|
florian
|
116 |
#define PRIbig "llu" // printf format for big_t |
|
6ea30fb…
|
florian
|
117 |
typedef big_t code_t; // type for bit pattern counting |
|
adb9e8e…
|
drh
|
118 |
struct tab { // type for been-here check |
|
adb9e8e…
|
drh
|
119 |
size_t len; // allocated length of bit vector in octets |
|
adb9e8e…
|
drh
|
120 |
char *vec; // allocated bit vector |
|
7ef7284…
|
drh
|
121 |
}; |
|
7ef7284…
|
drh
|
122 |
|
|
7ef7284…
|
drh
|
123 |
/* The array for saving results, num[], is indexed with this triplet: |
|
7ef7284…
|
drh
|
124 |
|
|
7ef7284…
|
drh
|
125 |
syms: number of symbols remaining to code |
|
7ef7284…
|
drh
|
126 |
left: number of available bit patterns at length len |
|
7ef7284…
|
drh
|
127 |
len: number of bits in the codes currently being assigned |
|
7ef7284…
|
drh
|
128 |
|
|
7ef7284…
|
drh
|
129 |
Those indices are constrained thusly when saving results: |
|
7ef7284…
|
drh
|
130 |
|
|
7ef7284…
|
drh
|
131 |
syms: 3..totsym (totsym == total symbols to code) |
|
7ef7284…
|
drh
|
132 |
left: 2..syms - 1, but only the evens (so syms == 8 -> 2, 4, 6) |
|
7ef7284…
|
drh
|
133 |
len: 1..max - 1 (max == maximum code length in bits) |
|
7ef7284…
|
drh
|
134 |
|
|
adb9e8e…
|
drh
|
135 |
syms == 2 is not saved since that immediately leads to a single code. left |
|
7ef7284…
|
drh
|
136 |
must be even, since it represents the number of available bit patterns at |
|
adb9e8e…
|
drh
|
137 |
the current length, which is double the number at the previous length. left |
|
adb9e8e…
|
drh
|
138 |
ends at syms-1 since left == syms immediately results in a single code. |
|
7ef7284…
|
drh
|
139 |
(left > sym is not allowed since that would result in an incomplete code.) |
|
7ef7284…
|
drh
|
140 |
len is less than max, since the code completes immediately when len == max. |
|
7ef7284…
|
drh
|
141 |
|
|
adb9e8e…
|
drh
|
142 |
The offset into the array is calculated for the three indices with the first |
|
adb9e8e…
|
drh
|
143 |
one (syms) being outermost, and the last one (len) being innermost. We build |
|
adb9e8e…
|
drh
|
144 |
the array with length max-1 lists for the len index, with syms-3 of those |
|
adb9e8e…
|
drh
|
145 |
for each symbol. There are totsym-2 of those, with each one varying in |
|
adb9e8e…
|
drh
|
146 |
length as a function of sym. See the calculation of index in map() for the |
|
adb9e8e…
|
drh
|
147 |
index, and the calculation of size in main() for the size of the array. |
|
7ef7284…
|
drh
|
148 |
|
|
7ef7284…
|
drh
|
149 |
For the deflate example of 286 symbols limited to 15-bit codes, the array |
|
adb9e8e…
|
drh
|
150 |
has 284,284 entries, taking up 2.17 MB for an 8-byte big_t. More than half |
|
adb9e8e…
|
drh
|
151 |
of the space allocated for saved results is actually used -- not all |
|
adb9e8e…
|
drh
|
152 |
possible triplets are reached in the generation of valid prefix codes. |
|
7ef7284…
|
drh
|
153 |
*/ |
|
7ef7284…
|
drh
|
154 |
|
|
7ef7284…
|
drh
|
155 |
/* The array for tracking visited states, done[], is itself indexed identically |
|
7ef7284…
|
drh
|
156 |
to the num[] array as described above for the (syms, left, len) triplet. |
|
7ef7284…
|
drh
|
157 |
Each element in the array is further indexed by the (mem, rem) doublet, |
|
7ef7284…
|
drh
|
158 |
where mem is the amount of inflate table space used so far, and rem is the |
|
adb9e8e…
|
drh
|
159 |
remaining unused entries in the current inflate sub-table. Each indexed |
|
7ef7284…
|
drh
|
160 |
element is simply one bit indicating whether the state has been visited or |
|
adb9e8e…
|
drh
|
161 |
not. Since the ranges for mem and rem are not known a priori, each bit |
|
7ef7284…
|
drh
|
162 |
vector is of a variable size, and grows as needed to accommodate the visited |
|
adb9e8e…
|
drh
|
163 |
states. mem and rem are used to calculate a single index in a triangular |
|
adb9e8e…
|
drh
|
164 |
array. Since the range of mem is expected in the default case to be about |
|
7ef7284…
|
drh
|
165 |
ten times larger than the range of rem, the array is skewed to reduce the |
|
adb9e8e…
|
drh
|
166 |
memory usage, with eight times the range for mem than for rem. See the |
|
adb9e8e…
|
drh
|
167 |
calculations for offset and bit in been_here() for the details. |
|
7ef7284…
|
drh
|
168 |
|
|
7ef7284…
|
drh
|
169 |
For the deflate example of 286 symbols limited to 15-bit codes, the bit |
|
adb9e8e…
|
drh
|
170 |
vectors grow to total 5.5 MB, in addition to the 4.3 MB done array itself. |
|
7ef7284…
|
drh
|
171 |
*/ |
|
7ef7284…
|
drh
|
172 |
|
|
adb9e8e…
|
drh
|
173 |
// Type for a variable-length, allocated string. |
|
adb9e8e…
|
drh
|
174 |
typedef struct { |
|
adb9e8e…
|
drh
|
175 |
char *str; // pointer to allocated string |
|
adb9e8e…
|
drh
|
176 |
size_t size; // size of allocation |
|
adb9e8e…
|
drh
|
177 |
size_t len; // length of string, not including terminating zero |
|
adb9e8e…
|
drh
|
178 |
} string_t; |
|
adb9e8e…
|
drh
|
179 |
|
|
adb9e8e…
|
drh
|
180 |
// Clear a string_t. |
|
adb9e8e…
|
drh
|
181 |
local void string_clear(string_t *s) { |
|
adb9e8e…
|
drh
|
182 |
s->str[0] = 0; |
|
adb9e8e…
|
drh
|
183 |
s->len = 0; |
|
adb9e8e…
|
drh
|
184 |
} |
|
adb9e8e…
|
drh
|
185 |
|
|
adb9e8e…
|
drh
|
186 |
// Initialize a string_t. |
|
adb9e8e…
|
drh
|
187 |
local void string_init(string_t *s) { |
|
adb9e8e…
|
drh
|
188 |
s->size = 16; |
|
adb9e8e…
|
drh
|
189 |
s->str = malloc(s->size); |
|
adb9e8e…
|
drh
|
190 |
assert(s->str != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
191 |
string_clear(s); |
|
adb9e8e…
|
drh
|
192 |
} |
|
adb9e8e…
|
drh
|
193 |
|
|
adb9e8e…
|
drh
|
194 |
// Release the allocation of a string_t. |
|
adb9e8e…
|
drh
|
195 |
local void string_free(string_t *s) { |
|
adb9e8e…
|
drh
|
196 |
free(s->str); |
|
adb9e8e…
|
drh
|
197 |
s->str = NULL; |
|
adb9e8e…
|
drh
|
198 |
s->size = 0; |
|
adb9e8e…
|
drh
|
199 |
s->len = 0; |
|
adb9e8e…
|
drh
|
200 |
} |
|
adb9e8e…
|
drh
|
201 |
|
|
adb9e8e…
|
drh
|
202 |
// Save the results of printf with fmt and the subsequent argument list to s. |
|
adb9e8e…
|
drh
|
203 |
// Each call appends to s. The allocated space for s is increased as needed. |
|
adb9e8e…
|
drh
|
204 |
local void string_printf(string_t *s, char *fmt, ...) { |
|
adb9e8e…
|
drh
|
205 |
va_list ap; |
|
adb9e8e…
|
drh
|
206 |
va_start(ap, fmt); |
|
adb9e8e…
|
drh
|
207 |
size_t len = s->len; |
|
adb9e8e…
|
drh
|
208 |
int ret = vsnprintf(s->str + len, s->size - len, fmt, ap); |
|
adb9e8e…
|
drh
|
209 |
assert(ret >= 0 && "out of memory"); |
|
adb9e8e…
|
drh
|
210 |
s->len += ret; |
|
adb9e8e…
|
drh
|
211 |
if (s->size < s->len + 1) { |
|
adb9e8e…
|
drh
|
212 |
do { |
|
adb9e8e…
|
drh
|
213 |
s->size <<= 1; |
|
adb9e8e…
|
drh
|
214 |
assert(s->size != 0 && "overflow"); |
|
adb9e8e…
|
drh
|
215 |
} while (s->size < s->len + 1); |
|
adb9e8e…
|
drh
|
216 |
s->str = realloc(s->str, s->size); |
|
adb9e8e…
|
drh
|
217 |
assert(s->str != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
218 |
vsnprintf(s->str + len, s->size - len, fmt, ap); |
|
adb9e8e…
|
drh
|
219 |
} |
|
adb9e8e…
|
drh
|
220 |
va_end(ap); |
|
adb9e8e…
|
drh
|
221 |
} |
|
adb9e8e…
|
drh
|
222 |
|
|
adb9e8e…
|
drh
|
223 |
// Globals to avoid propagating constants or constant pointers recursively. |
|
adb9e8e…
|
drh
|
224 |
struct { |
|
adb9e8e…
|
drh
|
225 |
int max; // maximum allowed bit length for the codes |
|
adb9e8e…
|
drh
|
226 |
int root; // size of base code table in bits |
|
adb9e8e…
|
drh
|
227 |
int large; // largest code table so far |
|
adb9e8e…
|
drh
|
228 |
size_t size; // number of elements in num and done |
|
adb9e8e…
|
drh
|
229 |
big_t tot; // total number of codes with maximum tables size |
|
adb9e8e…
|
drh
|
230 |
string_t out; // display of subcodes for maximum tables size |
|
adb9e8e…
|
drh
|
231 |
int *code; // number of symbols assigned to each bit length |
|
adb9e8e…
|
drh
|
232 |
big_t *num; // saved results array for code counting |
|
adb9e8e…
|
drh
|
233 |
struct tab *done; // states already evaluated array |
|
adb9e8e…
|
drh
|
234 |
} g; |
|
adb9e8e…
|
drh
|
235 |
|
|
adb9e8e…
|
drh
|
236 |
// Index function for num[] and done[]. |
|
adb9e8e…
|
drh
|
237 |
local inline size_t map(int syms, int left, int len) { |
|
adb9e8e…
|
drh
|
238 |
return ((size_t)((syms - 1) >> 1) * ((syms - 2) >> 1) + |
|
adb9e8e…
|
drh
|
239 |
(left >> 1) - 1) * (g.max - 1) + |
|
adb9e8e…
|
drh
|
240 |
len - 1; |
|
adb9e8e…
|
drh
|
241 |
} |
|
adb9e8e…
|
drh
|
242 |
|
|
adb9e8e…
|
drh
|
243 |
// Free allocated space in globals. |
|
adb9e8e…
|
drh
|
244 |
local void cleanup(void) { |
|
adb9e8e…
|
drh
|
245 |
if (g.done != NULL) { |
|
adb9e8e…
|
drh
|
246 |
for (size_t n = 0; n < g.size; n++) |
|
adb9e8e…
|
drh
|
247 |
if (g.done[n].len) |
|
adb9e8e…
|
drh
|
248 |
free(g.done[n].vec); |
|
adb9e8e…
|
drh
|
249 |
g.size = 0; |
|
adb9e8e…
|
drh
|
250 |
free(g.done); g.done = NULL; |
|
adb9e8e…
|
drh
|
251 |
} |
|
adb9e8e…
|
drh
|
252 |
free(g.num); g.num = NULL; |
|
adb9e8e…
|
drh
|
253 |
free(g.code); g.code = NULL; |
|
adb9e8e…
|
drh
|
254 |
string_free(&g.out); |
|
adb9e8e…
|
drh
|
255 |
} |
|
adb9e8e…
|
drh
|
256 |
|
|
adb9e8e…
|
drh
|
257 |
// Return the number of possible prefix codes using bit patterns of lengths len |
|
adb9e8e…
|
drh
|
258 |
// through max inclusive, coding syms symbols, with left bit patterns of length |
|
adb9e8e…
|
drh
|
259 |
// len unused -- return -1 if there is an overflow in the counting. Keep a |
|
adb9e8e…
|
drh
|
260 |
// record of previous results in num to prevent repeating the same calculation. |
|
adb9e8e…
|
drh
|
261 |
local big_t count(int syms, int left, int len) { |
|
adb9e8e…
|
drh
|
262 |
// see if only one possible code |
|
7ef7284…
|
drh
|
263 |
if (syms == left) |
|
7ef7284…
|
drh
|
264 |
return 1; |
|
7ef7284…
|
drh
|
265 |
|
|
adb9e8e…
|
drh
|
266 |
// note and verify the expected state |
|
adb9e8e…
|
drh
|
267 |
assert(syms > left && left > 0 && len < g.max); |
|
7ef7284…
|
drh
|
268 |
|
|
adb9e8e…
|
drh
|
269 |
// see if we've done this one already |
|
adb9e8e…
|
drh
|
270 |
size_t index = map(syms, left, len); |
|
adb9e8e…
|
drh
|
271 |
big_t got = g.num[index]; |
|
7ef7284…
|
drh
|
272 |
if (got) |
|
adb9e8e…
|
drh
|
273 |
return got; // we have -- return the saved result |
|
7ef7284…
|
drh
|
274 |
|
|
adb9e8e…
|
drh
|
275 |
// we need to use at least this many bit patterns so that the code won't be |
|
adb9e8e…
|
drh
|
276 |
// incomplete at the next length (more bit patterns than symbols) |
|
adb9e8e…
|
drh
|
277 |
int least = (left << 1) - syms; |
|
7ef7284…
|
drh
|
278 |
if (least < 0) |
|
7ef7284…
|
drh
|
279 |
least = 0; |
|
7ef7284…
|
drh
|
280 |
|
|
adb9e8e…
|
drh
|
281 |
// we can use at most this many bit patterns, lest there not be enough |
|
adb9e8e…
|
drh
|
282 |
// available for the remaining symbols at the maximum length (if there were |
|
adb9e8e…
|
drh
|
283 |
// no limit to the code length, this would become: most = left - 1) |
|
adb9e8e…
|
drh
|
284 |
int most = (((code_t)left << (g.max - len)) - syms) / |
|
adb9e8e…
|
drh
|
285 |
(((code_t)1 << (g.max - len)) - 1); |
|
adb9e8e…
|
drh
|
286 |
|
|
adb9e8e…
|
drh
|
287 |
// count all possible codes from this juncture and add them up |
|
adb9e8e…
|
drh
|
288 |
big_t sum = 0; |
|
adb9e8e…
|
drh
|
289 |
for (int use = least; use <= most; use++) { |
|
adb9e8e…
|
drh
|
290 |
got = count(syms - use, (left - use) << 1, len + 1); |
|
7ef7284…
|
drh
|
291 |
sum += got; |
|
adb9e8e…
|
drh
|
292 |
if (got == (big_t)-1 || sum < got) // overflow |
|
adb9e8e…
|
drh
|
293 |
return (big_t)-1; |
|
7ef7284…
|
drh
|
294 |
} |
|
7ef7284…
|
drh
|
295 |
|
|
adb9e8e…
|
drh
|
296 |
// verify that all recursive calls are productive |
|
7ef7284…
|
drh
|
297 |
assert(sum != 0); |
|
7ef7284…
|
drh
|
298 |
|
|
adb9e8e…
|
drh
|
299 |
// save the result and return it |
|
adb9e8e…
|
drh
|
300 |
g.num[index] = sum; |
|
7ef7284…
|
drh
|
301 |
return sum; |
|
7ef7284…
|
drh
|
302 |
} |
|
7ef7284…
|
drh
|
303 |
|
|
adb9e8e…
|
drh
|
304 |
// Return true if we've been here before, set to true if not. Set a bit in a |
|
adb9e8e…
|
drh
|
305 |
// bit vector to indicate visiting this state. Each (syms,len,left) state has a |
|
adb9e8e…
|
drh
|
306 |
// variable size bit vector indexed by (mem,rem). The bit vector is lengthened |
|
adb9e8e…
|
drh
|
307 |
// as needed to allow setting the (mem,rem) bit. |
|
adb9e8e…
|
drh
|
308 |
local int been_here(int syms, int left, int len, int mem, int rem) { |
|
adb9e8e…
|
drh
|
309 |
// point to vector for (syms,left,len), bit in vector for (mem,rem) |
|
adb9e8e…
|
drh
|
310 |
size_t index = map(syms, left, len); |
|
adb9e8e…
|
drh
|
311 |
mem -= 1 << g.root; // mem always includes the root table |
|
adb9e8e…
|
drh
|
312 |
mem >>= 1; // mem and rem are always even |
|
adb9e8e…
|
drh
|
313 |
rem >>= 1; |
|
adb9e8e…
|
drh
|
314 |
size_t offset = (mem >> 3) + rem; |
|
7ef7284…
|
drh
|
315 |
offset = ((offset * (offset + 1)) >> 1) + rem; |
|
adb9e8e…
|
drh
|
316 |
int bit = 1 << (mem & 7); |
|
adb9e8e…
|
drh
|
317 |
|
|
adb9e8e…
|
drh
|
318 |
// see if we've been here |
|
adb9e8e…
|
drh
|
319 |
size_t length = g.done[index].len; |
|
adb9e8e…
|
drh
|
320 |
if (offset < length && (g.done[index].vec[offset] & bit) != 0) |
|
adb9e8e…
|
drh
|
321 |
return 1; // done this! |
|
adb9e8e…
|
drh
|
322 |
|
|
adb9e8e…
|
drh
|
323 |
// we haven't been here before -- set the bit to show we have now |
|
adb9e8e…
|
drh
|
324 |
|
|
adb9e8e…
|
drh
|
325 |
// see if we need to lengthen the vector in order to set the bit |
|
7ef7284…
|
drh
|
326 |
if (length <= offset) { |
|
adb9e8e…
|
drh
|
327 |
// if we have one already, enlarge it, zero out the appended space |
|
adb9e8e…
|
drh
|
328 |
char *vector; |
|
7ef7284…
|
drh
|
329 |
if (length) { |
|
7ef7284…
|
drh
|
330 |
do { |
|
7ef7284…
|
drh
|
331 |
length <<= 1; |
|
7ef7284…
|
drh
|
332 |
} while (length <= offset); |
|
adb9e8e…
|
drh
|
333 |
vector = realloc(g.done[index].vec, length); |
|
adb9e8e…
|
drh
|
334 |
assert(vector != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
335 |
memset(vector + g.done[index].len, 0, length - g.done[index].len); |
|
adb9e8e…
|
drh
|
336 |
} |
|
adb9e8e…
|
drh
|
337 |
|
|
adb9e8e…
|
drh
|
338 |
// otherwise we need to make a new vector and zero it out |
|
adb9e8e…
|
drh
|
339 |
else { |
|
adb9e8e…
|
drh
|
340 |
length = 16; |
|
adb9e8e…
|
drh
|
341 |
while (length <= offset) |
|
adb9e8e…
|
drh
|
342 |
length <<= 1; |
|
adb9e8e…
|
drh
|
343 |
vector = calloc(length, 1); |
|
adb9e8e…
|
drh
|
344 |
assert(vector != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
345 |
} |
|
adb9e8e…
|
drh
|
346 |
|
|
adb9e8e…
|
drh
|
347 |
// install the new vector |
|
adb9e8e…
|
drh
|
348 |
g.done[index].len = length; |
|
adb9e8e…
|
drh
|
349 |
g.done[index].vec = vector; |
|
adb9e8e…
|
drh
|
350 |
} |
|
adb9e8e…
|
drh
|
351 |
|
|
adb9e8e…
|
drh
|
352 |
// set the bit |
|
adb9e8e…
|
drh
|
353 |
g.done[index].vec[offset] |= bit; |
|
7ef7284…
|
drh
|
354 |
return 0; |
|
7ef7284…
|
drh
|
355 |
} |
|
7ef7284…
|
drh
|
356 |
|
|
adb9e8e…
|
drh
|
357 |
// Examine all possible codes from the given node (syms, len, left). Compute |
|
adb9e8e…
|
drh
|
358 |
// the amount of memory required to build inflate's decoding tables, where the |
|
adb9e8e…
|
drh
|
359 |
// number of code structures used so far is mem, and the number remaining in |
|
adb9e8e…
|
drh
|
360 |
// the current sub-table is rem. |
|
adb9e8e…
|
drh
|
361 |
local void examine(int syms, int left, int len, int mem, int rem) { |
|
adb9e8e…
|
drh
|
362 |
// see if we have a complete code |
|
7ef7284…
|
drh
|
363 |
if (syms == left) { |
|
adb9e8e…
|
drh
|
364 |
// set the last code entry |
|
adb9e8e…
|
drh
|
365 |
g.code[len] = left; |
|
7ef7284…
|
drh
|
366 |
|
|
adb9e8e…
|
drh
|
367 |
// complete computation of memory used by this code |
|
7ef7284…
|
drh
|
368 |
while (rem < left) { |
|
7ef7284…
|
drh
|
369 |
left -= rem; |
|
adb9e8e…
|
drh
|
370 |
rem = 1 << (len - g.root); |
|
7ef7284…
|
drh
|
371 |
mem += rem; |
|
7ef7284…
|
drh
|
372 |
} |
|
7ef7284…
|
drh
|
373 |
assert(rem == left); |
|
7ef7284…
|
drh
|
374 |
|
|
adb9e8e…
|
drh
|
375 |
// if this is at the maximum, show the sub-code |
|
adb9e8e…
|
drh
|
376 |
if (mem >= g.large) { |
|
adb9e8e…
|
drh
|
377 |
// if this is a new maximum, update the maximum and clear out the |
|
adb9e8e…
|
drh
|
378 |
// printed sub-codes from the previous maximum |
|
adb9e8e…
|
drh
|
379 |
if (mem > g.large) { |
|
adb9e8e…
|
drh
|
380 |
g.large = mem; |
|
adb9e8e…
|
drh
|
381 |
string_clear(&g.out); |
|
adb9e8e…
|
drh
|
382 |
} |
|
adb9e8e…
|
drh
|
383 |
|
|
adb9e8e…
|
drh
|
384 |
// compute the starting state for this sub-code |
|
adb9e8e…
|
drh
|
385 |
syms = 0; |
|
adb9e8e…
|
drh
|
386 |
left = 1 << g.max; |
|
adb9e8e…
|
drh
|
387 |
for (int bits = g.max; bits > g.root; bits--) { |
|
adb9e8e…
|
drh
|
388 |
syms += g.code[bits]; |
|
adb9e8e…
|
drh
|
389 |
left -= g.code[bits]; |
|
adb9e8e…
|
drh
|
390 |
assert((left & 1) == 0); |
|
adb9e8e…
|
drh
|
391 |
left >>= 1; |
|
adb9e8e…
|
drh
|
392 |
} |
|
adb9e8e…
|
drh
|
393 |
|
|
adb9e8e…
|
drh
|
394 |
// print the starting state and the resulting sub-code to g.out |
|
adb9e8e…
|
drh
|
395 |
string_printf(&g.out, "<%u, %u, %u>:", |
|
adb9e8e…
|
drh
|
396 |
syms, g.root + 1, ((1 << g.root) - left) << 1); |
|
adb9e8e…
|
drh
|
397 |
for (int bits = g.root + 1; bits <= g.max; bits++) |
|
adb9e8e…
|
drh
|
398 |
if (g.code[bits]) |
|
adb9e8e…
|
drh
|
399 |
string_printf(&g.out, " %d[%d]", g.code[bits], bits); |
|
adb9e8e…
|
drh
|
400 |
string_printf(&g.out, "\n"); |
|
adb9e8e…
|
drh
|
401 |
} |
|
adb9e8e…
|
drh
|
402 |
|
|
adb9e8e…
|
drh
|
403 |
// remove entries as we drop back down in the recursion |
|
adb9e8e…
|
drh
|
404 |
g.code[len] = 0; |
|
adb9e8e…
|
drh
|
405 |
return; |
|
adb9e8e…
|
drh
|
406 |
} |
|
adb9e8e…
|
drh
|
407 |
|
|
adb9e8e…
|
drh
|
408 |
// prune the tree if we can |
|
adb9e8e…
|
drh
|
409 |
if (been_here(syms, left, len, mem, rem)) |
|
adb9e8e…
|
drh
|
410 |
return; |
|
adb9e8e…
|
drh
|
411 |
|
|
adb9e8e…
|
drh
|
412 |
// we need to use at least this many bit patterns so that the code won't be |
|
adb9e8e…
|
drh
|
413 |
// incomplete at the next length (more bit patterns than symbols) |
|
adb9e8e…
|
drh
|
414 |
int least = (left << 1) - syms; |
|
7ef7284…
|
drh
|
415 |
if (least < 0) |
|
7ef7284…
|
drh
|
416 |
least = 0; |
|
7ef7284…
|
drh
|
417 |
|
|
adb9e8e…
|
drh
|
418 |
// we can use at most this many bit patterns, lest there not be enough |
|
adb9e8e…
|
drh
|
419 |
// available for the remaining symbols at the maximum length (if there were |
|
adb9e8e…
|
drh
|
420 |
// no limit to the code length, this would become: most = left - 1) |
|
adb9e8e…
|
drh
|
421 |
int most = (((code_t)left << (g.max - len)) - syms) / |
|
adb9e8e…
|
drh
|
422 |
(((code_t)1 << (g.max - len)) - 1); |
|
adb9e8e…
|
drh
|
423 |
|
|
adb9e8e…
|
drh
|
424 |
// occupy least table spaces, creating new sub-tables as needed |
|
adb9e8e…
|
drh
|
425 |
int use = least; |
|
7ef7284…
|
drh
|
426 |
while (rem < use) { |
|
7ef7284…
|
drh
|
427 |
use -= rem; |
|
adb9e8e…
|
drh
|
428 |
rem = 1 << (len - g.root); |
|
7ef7284…
|
drh
|
429 |
mem += rem; |
|
7ef7284…
|
drh
|
430 |
} |
|
7ef7284…
|
drh
|
431 |
rem -= use; |
|
7ef7284…
|
drh
|
432 |
|
|
adb9e8e…
|
drh
|
433 |
// examine codes from here, updating table space as we go |
|
7ef7284…
|
drh
|
434 |
for (use = least; use <= most; use++) { |
|
adb9e8e…
|
drh
|
435 |
g.code[len] = use; |
|
adb9e8e…
|
drh
|
436 |
examine(syms - use, (left - use) << 1, len + 1, |
|
adb9e8e…
|
drh
|
437 |
mem + (rem ? 1 << (len - g.root) : 0), rem << 1); |
|
7ef7284…
|
drh
|
438 |
if (rem == 0) { |
|
adb9e8e…
|
drh
|
439 |
rem = 1 << (len - g.root); |
|
7ef7284…
|
drh
|
440 |
mem += rem; |
|
7ef7284…
|
drh
|
441 |
} |
|
7ef7284…
|
drh
|
442 |
rem--; |
|
7ef7284…
|
drh
|
443 |
} |
|
7ef7284…
|
drh
|
444 |
|
|
adb9e8e…
|
drh
|
445 |
// remove entries as we drop back down in the recursion |
|
adb9e8e…
|
drh
|
446 |
g.code[len] = 0; |
|
adb9e8e…
|
drh
|
447 |
} |
|
adb9e8e…
|
drh
|
448 |
|
|
adb9e8e…
|
drh
|
449 |
// Look at all sub-codes starting with root + 1 bits. Look at only the valid |
|
adb9e8e…
|
drh
|
450 |
// intermediate code states (syms, left, len). For each completed code, |
|
adb9e8e…
|
drh
|
451 |
// calculate the amount of memory required by inflate to build the decoding |
|
adb9e8e…
|
drh
|
452 |
// tables. Find the maximum amount of memory required and show the codes that |
|
adb9e8e…
|
drh
|
453 |
// require that maximum. |
|
adb9e8e…
|
drh
|
454 |
local void enough(int syms) { |
|
adb9e8e…
|
drh
|
455 |
// clear code |
|
adb9e8e…
|
drh
|
456 |
for (int n = 0; n <= g.max; n++) |
|
adb9e8e…
|
drh
|
457 |
g.code[n] = 0; |
|
adb9e8e…
|
drh
|
458 |
|
|
adb9e8e…
|
drh
|
459 |
// look at all (root + 1) bit and longer codes |
|
adb9e8e…
|
drh
|
460 |
string_clear(&g.out); // empty saved results |
|
adb9e8e…
|
drh
|
461 |
g.large = 1 << g.root; // base table |
|
adb9e8e…
|
drh
|
462 |
if (g.root < g.max) // otherwise, there's only a base table |
|
adb9e8e…
|
drh
|
463 |
for (int n = 3; n <= syms; n++) |
|
adb9e8e…
|
drh
|
464 |
for (int left = 2; left < n; left += 2) { |
|
adb9e8e…
|
drh
|
465 |
// look at all reachable (root + 1) bit nodes, and the |
|
adb9e8e…
|
drh
|
466 |
// resulting codes (complete at root + 2 or more) |
|
adb9e8e…
|
drh
|
467 |
size_t index = map(n, left, g.root + 1); |
|
adb9e8e…
|
drh
|
468 |
if (g.root + 1 < g.max && g.num[index]) // reachable node |
|
adb9e8e…
|
drh
|
469 |
examine(n, left, g.root + 1, 1 << g.root, 0); |
|
adb9e8e…
|
drh
|
470 |
|
|
adb9e8e…
|
drh
|
471 |
// also look at root bit codes with completions at root + 1 |
|
adb9e8e…
|
drh
|
472 |
// bits (not saved in num, since complete), just in case |
|
adb9e8e…
|
drh
|
473 |
if (g.num[index - 1] && n <= left << 1) |
|
adb9e8e…
|
drh
|
474 |
examine((n - left) << 1, (n - left) << 1, g.root + 1, |
|
adb9e8e…
|
drh
|
475 |
1 << g.root, 0); |
|
adb9e8e…
|
drh
|
476 |
} |
|
adb9e8e…
|
drh
|
477 |
|
|
adb9e8e…
|
drh
|
478 |
// done |
|
adb9e8e…
|
drh
|
479 |
printf("maximum of %d table entries for root = %d\n", g.large, g.root); |
|
adb9e8e…
|
drh
|
480 |
fputs(g.out.str, stdout); |
|
adb9e8e…
|
drh
|
481 |
} |
|
adb9e8e…
|
drh
|
482 |
|
|
adb9e8e…
|
drh
|
483 |
// Examine and show the total number of possible prefix codes for a given |
|
adb9e8e…
|
drh
|
484 |
// maximum number of symbols, initial root table size, and maximum code length |
|
adb9e8e…
|
drh
|
485 |
// in bits -- those are the command arguments in that order. The default values |
|
adb9e8e…
|
drh
|
486 |
// are 286, 9, and 15 respectively, for the deflate literal/length code. The |
|
adb9e8e…
|
drh
|
487 |
// possible codes are counted for each number of coded symbols from two to the |
|
adb9e8e…
|
drh
|
488 |
// maximum. The counts for each of those and the total number of codes are |
|
a9e589c…
|
florian
|
489 |
// shown. The maximum number of inflate table entries is then calculated across |
|
adb9e8e…
|
drh
|
490 |
// all possible codes. Each new maximum number of table entries and the |
|
adb9e8e…
|
drh
|
491 |
// associated sub-code (starting at root + 1 == 10 bits) is shown. |
|
adb9e8e…
|
drh
|
492 |
// |
|
adb9e8e…
|
drh
|
493 |
// To count and examine prefix codes that are not length-limited, provide a |
|
adb9e8e…
|
drh
|
494 |
// maximum length equal to the number of symbols minus one. |
|
adb9e8e…
|
drh
|
495 |
// |
|
adb9e8e…
|
drh
|
496 |
// For the deflate literal/length code, use "enough". For the deflate distance |
|
adb9e8e…
|
drh
|
497 |
// code, use "enough 30 6". |
|
adb9e8e…
|
drh
|
498 |
int main(int argc, char **argv) { |
|
adb9e8e…
|
drh
|
499 |
// set up globals for cleanup() |
|
adb9e8e…
|
drh
|
500 |
g.code = NULL; |
|
adb9e8e…
|
drh
|
501 |
g.num = NULL; |
|
adb9e8e…
|
drh
|
502 |
g.done = NULL; |
|
adb9e8e…
|
drh
|
503 |
string_init(&g.out); |
|
adb9e8e…
|
drh
|
504 |
|
|
adb9e8e…
|
drh
|
505 |
// get arguments -- default to the deflate literal/length code |
|
adb9e8e…
|
drh
|
506 |
int syms = 286; |
|
adb9e8e…
|
drh
|
507 |
g.root = 9; |
|
adb9e8e…
|
drh
|
508 |
g.max = 15; |
|
7ef7284…
|
drh
|
509 |
if (argc > 1) { |
|
7ef7284…
|
drh
|
510 |
syms = atoi(argv[1]); |
|
7ef7284…
|
drh
|
511 |
if (argc > 2) { |
|
adb9e8e…
|
drh
|
512 |
g.root = atoi(argv[2]); |
|
bb4776e…
|
jan.nijtmans
|
513 |
if (argc > 3) |
|
adb9e8e…
|
drh
|
514 |
g.max = atoi(argv[3]); |
|
bb4776e…
|
jan.nijtmans
|
515 |
} |
|
7ef7284…
|
drh
|
516 |
} |
|
adb9e8e…
|
drh
|
517 |
if (argc > 4 || syms < 2 || g.root < 1 || g.max < 1) { |
|
7ef7284…
|
drh
|
518 |
fputs("invalid arguments, need: [sym >= 2 [root >= 1 [max >= 1]]]\n", |
|
bb4776e…
|
jan.nijtmans
|
519 |
stderr); |
|
7ef7284…
|
drh
|
520 |
return 1; |
|
7ef7284…
|
drh
|
521 |
} |
|
7ef7284…
|
drh
|
522 |
|
|
adb9e8e…
|
drh
|
523 |
// if not restricting the code length, the longest is syms - 1 |
|
adb9e8e…
|
drh
|
524 |
if (g.max > syms - 1) |
|
adb9e8e…
|
drh
|
525 |
g.max = syms - 1; |
|
adb9e8e…
|
drh
|
526 |
|
|
adb9e8e…
|
drh
|
527 |
// determine the number of bits in a code_t |
|
adb9e8e…
|
drh
|
528 |
int bits = 0; |
|
adb9e8e…
|
drh
|
529 |
for (code_t word = 1; word; word <<= 1) |
|
adb9e8e…
|
drh
|
530 |
bits++; |
|
adb9e8e…
|
drh
|
531 |
|
|
adb9e8e…
|
drh
|
532 |
// make sure that the calculation of most will not overflow |
|
adb9e8e…
|
drh
|
533 |
if (g.max > bits || (code_t)(syms - 2) >= ((code_t)-1 >> (g.max - 1))) { |
|
7ef7284…
|
drh
|
534 |
fputs("abort: code length too long for internal types\n", stderr); |
|
7ef7284…
|
drh
|
535 |
return 1; |
|
7ef7284…
|
drh
|
536 |
} |
|
7ef7284…
|
drh
|
537 |
|
|
adb9e8e…
|
drh
|
538 |
// reject impossible code requests |
|
adb9e8e…
|
drh
|
539 |
if ((code_t)(syms - 1) > ((code_t)1 << g.max) - 1) { |
|
adb9e8e…
|
drh
|
540 |
fprintf(stderr, "%d symbols cannot be coded in %d bits\n", |
|
adb9e8e…
|
drh
|
541 |
syms, g.max); |
|
adb9e8e…
|
drh
|
542 |
return 1; |
|
adb9e8e…
|
drh
|
543 |
} |
|
adb9e8e…
|
drh
|
544 |
|
|
adb9e8e…
|
drh
|
545 |
// allocate code vector |
|
adb9e8e…
|
drh
|
546 |
g.code = calloc(g.max + 1, sizeof(int)); |
|
adb9e8e…
|
drh
|
547 |
assert(g.code != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
548 |
|
|
adb9e8e…
|
drh
|
549 |
// determine size of saved results array, checking for overflows, |
|
adb9e8e…
|
drh
|
550 |
// allocate and clear the array (set all to zero with calloc()) |
|
adb9e8e…
|
drh
|
551 |
if (syms == 2) // iff max == 1 |
|
adb9e8e…
|
drh
|
552 |
g.num = NULL; // won't be saving any results |
|
adb9e8e…
|
drh
|
553 |
else { |
|
adb9e8e…
|
drh
|
554 |
g.size = syms >> 1; |
|
adb9e8e…
|
drh
|
555 |
int n = (syms - 1) >> 1; |
|
adb9e8e…
|
drh
|
556 |
assert(g.size <= (size_t)-1 / n && "overflow"); |
|
adb9e8e…
|
drh
|
557 |
g.size *= n; |
|
adb9e8e…
|
drh
|
558 |
n = g.max - 1; |
|
adb9e8e…
|
drh
|
559 |
assert(g.size <= (size_t)-1 / n && "overflow"); |
|
adb9e8e…
|
drh
|
560 |
g.size *= n; |
|
adb9e8e…
|
drh
|
561 |
g.num = calloc(g.size, sizeof(big_t)); |
|
adb9e8e…
|
drh
|
562 |
assert(g.num != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
563 |
} |
|
adb9e8e…
|
drh
|
564 |
|
|
adb9e8e…
|
drh
|
565 |
// count possible codes for all numbers of symbols, add up counts |
|
adb9e8e…
|
drh
|
566 |
big_t sum = 0; |
|
adb9e8e…
|
drh
|
567 |
for (int n = 2; n <= syms; n++) { |
|
adb9e8e…
|
drh
|
568 |
big_t got = count(n, 2, 1); |
|
adb9e8e…
|
drh
|
569 |
sum += got; |
|
adb9e8e…
|
drh
|
570 |
assert(got != (big_t)-1 && sum >= got && "overflow"); |
|
adb9e8e…
|
drh
|
571 |
} |
|
adb9e8e…
|
drh
|
572 |
printf("%"PRIbig" total codes for 2 to %d symbols", sum, syms); |
|
adb9e8e…
|
drh
|
573 |
if (g.max < syms - 1) |
|
adb9e8e…
|
drh
|
574 |
printf(" (%d-bit length limit)\n", g.max); |
|
7ef7284…
|
drh
|
575 |
else |
|
7ef7284…
|
drh
|
576 |
puts(" (no length limit)"); |
|
7ef7284…
|
drh
|
577 |
|
|
adb9e8e…
|
drh
|
578 |
// allocate and clear done array for been_here() |
|
adb9e8e…
|
drh
|
579 |
if (syms == 2) |
|
adb9e8e…
|
drh
|
580 |
g.done = NULL; |
|
adb9e8e…
|
drh
|
581 |
else { |
|
adb9e8e…
|
drh
|
582 |
g.done = calloc(g.size, sizeof(struct tab)); |
|
adb9e8e…
|
drh
|
583 |
assert(g.done != NULL && "out of memory"); |
|
adb9e8e…
|
drh
|
584 |
} |
|
adb9e8e…
|
drh
|
585 |
|
|
adb9e8e…
|
drh
|
586 |
// find and show maximum inflate table usage |
|
adb9e8e…
|
drh
|
587 |
if (g.root > g.max) // reduce root to max length |
|
adb9e8e…
|
drh
|
588 |
g.root = g.max; |
|
adb9e8e…
|
drh
|
589 |
if ((code_t)syms < ((code_t)1 << (g.root + 1))) |
|
adb9e8e…
|
drh
|
590 |
enough(syms); |
|
adb9e8e…
|
drh
|
591 |
else |
|
adb9e8e…
|
drh
|
592 |
fputs("cannot handle minimum code lengths > root", stderr); |
|
adb9e8e…
|
drh
|
593 |
|
|
adb9e8e…
|
drh
|
594 |
// done |
|
7ef7284…
|
drh
|
595 |
cleanup(); |
|
7ef7284…
|
drh
|
596 |
return 0; |
|
7ef7284…
|
drh
|
597 |
} |