1 |
/* |
2 |
* Copyright (c) 2000-2007 Marc Alexander Lehmann <schmorp@schmorp.de> |
3 |
* |
4 |
* Redistribution and use in source and binary forms, with or without modifica- |
5 |
* tion, are permitted provided that the following conditions are met: |
6 |
* |
7 |
* 1. Redistributions of source code must retain the above copyright notice, |
8 |
* this list of conditions and the following disclaimer. |
9 |
* |
10 |
* 2. Redistributions in binary form must reproduce the above copyright |
11 |
* notice, this list of conditions and the following disclaimer in the |
12 |
* documentation and/or other materials provided with the distribution. |
13 |
* |
14 |
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED |
15 |
* WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER- |
16 |
* CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO |
17 |
* EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE- |
18 |
* CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
19 |
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; |
20 |
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, |
21 |
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- |
22 |
* ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED |
23 |
* OF THE POSSIBILITY OF SUCH DAMAGE. |
24 |
* |
25 |
* Alternatively, the contents of this file may be used under the terms of |
26 |
* the GNU General Public License ("GPL") version 2 or any later version, |
27 |
* in which case the provisions of the GPL are applicable instead of |
28 |
* the above. If you wish to allow the use of your version of this file |
29 |
* only under the terms of the GPL and not to allow others to use your |
30 |
* version of this file under the BSD license, indicate your decision |
31 |
* by deleting the provisions above and replace them with the notice |
32 |
* and other provisions required by the GPL. If you do not delete the |
33 |
* provisions above, a recipient may use your version of this file under |
34 |
* either the BSD or the GPL. |
35 |
*/ |
36 |
|
37 |
#include "lzfP.h" |
38 |
|
39 |
#define HSIZE (1 << (HLOG)) |
40 |
|
41 |
/* |
42 |
* don't play with this unless you benchmark! |
43 |
* decompression is not dependent on the hash function |
44 |
* the hashing function might seem strange, just believe me |
45 |
* it works ;) |
46 |
*/ |
47 |
#ifndef FRST |
48 |
# define FRST(p) (((p[0]) << 8) | p[1]) |
49 |
# define NEXT(v,p) (((v) << 8) | p[2]) |
50 |
# if ULTRA_FAST |
51 |
# define IDX(h) (((h >> (3*8 - HLOG)) - h ) & (HSIZE - 1)) |
52 |
# elif VERY_FAST |
53 |
# define IDX(h) (((h >> (3*8 - HLOG)) - h*5) & (HSIZE - 1)) |
54 |
# else |
55 |
# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1)) |
56 |
# endif |
57 |
/*# define IDX(h) ((ip[0] * 121 ^ ip[1] * 33 ^ ip[2] * 1) & (HSIZE-1))*/ |
58 |
#endif |
59 |
/* |
60 |
* IDX works because it is very similar to a multiplicative hash, e.g. |
61 |
* ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1)) |
62 |
* the latter is also quite fast on newer CPUs, and compresses similarly. |
63 |
* |
64 |
* the next one is also quite good, albeit slow ;) |
65 |
* (int)(cos(h & 0xffffff) * 1e6) |
66 |
*/ |
67 |
|
68 |
#if 0 |
69 |
/* original lzv-like hash function, much worse and thus slower */ |
70 |
# define FRST(p) (p[0] << 5) ^ p[1] |
71 |
# define NEXT(v,p) ((v) << 5) ^ p[2] |
72 |
# define IDX(h) ((h) & (HSIZE - 1)) |
73 |
#endif |
74 |
|
75 |
#define MAX_LIT (1 << 5) |
76 |
#define MAX_OFF (1 << 13) |
77 |
#define MAX_REF ((1 << 8) + (1 << 3)) |
78 |
|
79 |
#if __GNUC__ >= 3 |
80 |
# define expect(expr,value) __builtin_expect ((expr),(value)) |
81 |
# define inline inline |
82 |
#else |
83 |
# define expect(expr,value) (expr) |
84 |
# define inline static |
85 |
#endif |
86 |
|
87 |
#define expect_false(expr) expect ((expr) != 0, 0) |
88 |
#define expect_true(expr) expect ((expr) != 0, 1) |
89 |
|
90 |
/* |
91 |
* compressed format |
92 |
* |
93 |
* 000LLLLL <L+1> ; literal |
94 |
* LLLooooo oooooooo ; backref L |
95 |
* 111ooooo LLLLLLLL oooooooo ; backref L+7 |
96 |
* |
97 |
*/ |
98 |
|
99 |
unsigned int |
100 |
lzf_compress (const void *const in_data, unsigned int in_len, |
101 |
void *out_data, unsigned int out_len |
102 |
#if LZF_STATE_ARG |
103 |
, LZF_STATE htab |
104 |
#endif |
105 |
) |
106 |
{ |
107 |
#if !LZF_STATE_ARG |
108 |
LZF_STATE htab; |
109 |
#endif |
110 |
const u8 **hslot; |
111 |
const u8 *ip = (const u8 *)in_data; |
112 |
u8 *op = (u8 *)out_data; |
113 |
const u8 *in_end = ip + in_len; |
114 |
u8 *out_end = op + out_len; |
115 |
const u8 *ref; |
116 |
|
117 |
unsigned int hval; |
118 |
unsigned long off; |
119 |
int lit; |
120 |
|
121 |
if (!in_len || !out_len) |
122 |
return 0; |
123 |
|
124 |
#if INIT_HTAB |
125 |
memset (htab, 0, sizeof (htab)); |
126 |
# if 0 |
127 |
for (hslot = htab; hslot < htab + HSIZE; hslot++) |
128 |
*hslot++ = ip; |
129 |
# endif |
130 |
#endif |
131 |
|
132 |
lit = 0; op++; /* start run */ |
133 |
|
134 |
hval = FRST (ip); |
135 |
while (ip < in_end - 2) |
136 |
{ |
137 |
hval = NEXT (hval, ip); |
138 |
hslot = htab + IDX (hval); |
139 |
ref = *hslot; *hslot = ip; |
140 |
|
141 |
if (1 |
142 |
#if INIT_HTAB && !USE_MEMCPY |
143 |
&& ref < ip /* the next test will actually take care of this, but this is faster */ |
144 |
#endif |
145 |
&& (off = ip - ref - 1) < MAX_OFF |
146 |
&& ip + 4 < in_end |
147 |
&& ref > (u8 *)in_data |
148 |
#if STRICT_ALIGN |
149 |
&& ref[0] == ip[0] |
150 |
&& ref[1] == ip[1] |
151 |
&& ref[2] == ip[2] |
152 |
#else |
153 |
&& *(u16 *)ref == *(u16 *)ip |
154 |
&& ref[2] == ip[2] |
155 |
#endif |
156 |
) |
157 |
{ |
158 |
/* match found at *ref++ */ |
159 |
unsigned int len = 2; |
160 |
unsigned int maxlen = in_end - ip - len; |
161 |
maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; |
162 |
|
163 |
if (expect_false (op + 1 + 3 >= out_end)) |
164 |
return 0; |
165 |
|
166 |
op [- lit - 1] = lit - 1; /* stop run */ |
167 |
op -= !lit; /* undo run if length is zero */ |
168 |
|
169 |
for (;;) |
170 |
{ |
171 |
if (expect_true (maxlen > 16)) |
172 |
{ |
173 |
len++; if (ref [len] != ip [len]) break; |
174 |
len++; if (ref [len] != ip [len]) break; |
175 |
len++; if (ref [len] != ip [len]) break; |
176 |
len++; if (ref [len] != ip [len]) break; |
177 |
|
178 |
len++; if (ref [len] != ip [len]) break; |
179 |
len++; if (ref [len] != ip [len]) break; |
180 |
len++; if (ref [len] != ip [len]) break; |
181 |
len++; if (ref [len] != ip [len]) break; |
182 |
|
183 |
len++; if (ref [len] != ip [len]) break; |
184 |
len++; if (ref [len] != ip [len]) break; |
185 |
len++; if (ref [len] != ip [len]) break; |
186 |
len++; if (ref [len] != ip [len]) break; |
187 |
|
188 |
len++; if (ref [len] != ip [len]) break; |
189 |
len++; if (ref [len] != ip [len]) break; |
190 |
len++; if (ref [len] != ip [len]) break; |
191 |
len++; if (ref [len] != ip [len]) break; |
192 |
} |
193 |
|
194 |
do |
195 |
len++; |
196 |
while (len < maxlen && ref[len] == ip[len]); |
197 |
|
198 |
break; |
199 |
} |
200 |
|
201 |
len -= 2; |
202 |
ip++; |
203 |
|
204 |
if (len < 7) |
205 |
{ |
206 |
*op++ = (off >> 8) + (len << 5); |
207 |
} |
208 |
else |
209 |
{ |
210 |
*op++ = (off >> 8) + ( 7 << 5); |
211 |
*op++ = len - 7; |
212 |
} |
213 |
|
214 |
*op++ = off; |
215 |
|
216 |
#if ULTRA_FAST || VERY_FAST |
217 |
ip += len; |
218 |
#if VERY_FAST && !ULTRA_FAST |
219 |
--ip; |
220 |
#endif |
221 |
hval = FRST (ip); |
222 |
|
223 |
hval = NEXT (hval, ip); |
224 |
htab[IDX (hval)] = ip; |
225 |
ip++; |
226 |
|
227 |
#if VERY_FAST && !ULTRA_FAST |
228 |
hval = NEXT (hval, ip); |
229 |
htab[IDX (hval)] = ip; |
230 |
ip++; |
231 |
#endif |
232 |
#else |
233 |
do |
234 |
{ |
235 |
hval = NEXT (hval, ip); |
236 |
htab[IDX (hval)] = ip; |
237 |
ip++; |
238 |
} |
239 |
while (len--); |
240 |
#endif |
241 |
lit = 0; op++; /* start run */ |
242 |
continue; |
243 |
} |
244 |
|
245 |
/* one more literal byte we must copy */ |
246 |
|
247 |
if (expect_false (op >= out_end)) |
248 |
return 0; |
249 |
|
250 |
lit++; |
251 |
*op++ = *ip++; |
252 |
|
253 |
if (expect_false (lit == MAX_LIT)) |
254 |
{ |
255 |
op [- lit - 1] = lit - 1; /* stop run */ |
256 |
lit = 0; op++; /* start run */ |
257 |
} |
258 |
} |
259 |
|
260 |
if (op + lit + 2 >= out_end) |
261 |
return 0; |
262 |
|
263 |
while (ip < in_end) |
264 |
{ |
265 |
lit++; |
266 |
*op++ = *ip++; |
267 |
} |
268 |
|
269 |
op [- lit - 1] = lit - 1; |
270 |
|
271 |
return op - (u8 *)out_data; |
272 |
} |
273 |
|