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 |
# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1)) |
51 |
/*# define IDX(h) ((ip[0] * 121 ^ ip[1] * 33 ^ ip[2] * 1) & (HSIZE-1))*/ |
52 |
#endif |
53 |
/* |
54 |
* IDX works because it is very similar to a multiplicative hash, e.g. |
55 |
* ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1)) |
56 |
* the latter is also quite fast on newer CPUs, and compresses similarly. |
57 |
* |
58 |
* the next one is also quite good, albeit slow ;) |
59 |
* (int)(cos(h & 0xffffff) * 1e6) |
60 |
*/ |
61 |
|
62 |
#if 0 |
63 |
/* original lzv-like hash function, much worse and thus slower */ |
64 |
# define FRST(p) (p[0] << 5) ^ p[1] |
65 |
# define NEXT(v,p) ((v) << 5) ^ p[2] |
66 |
# define IDX(h) ((h) & (HSIZE - 1)) |
67 |
#endif |
68 |
|
69 |
#define MAX_LIT (1 << 5) |
70 |
#define MAX_OFF (1 << 13) |
71 |
#define MAX_REF ((1 << 8) + (1 << 3)) |
72 |
|
73 |
#if (__i386 || __amd64) && __GNUC__ >= 3 |
74 |
# define lzf_movsb(dst, src, len) \ |
75 |
asm ("rep movsb" \ |
76 |
: "=D" (dst), "=S" (src), "=c" (len) \ |
77 |
: "0" (dst), "1" (src), "2" (len)); |
78 |
#endif |
79 |
|
80 |
#if __GNUC__ >= 3 |
81 |
# define expect(expr,value) __builtin_expect ((expr),(value)) |
82 |
# define inline inline |
83 |
#else |
84 |
# define expect(expr,value) (expr) |
85 |
# define inline static |
86 |
#endif |
87 |
|
88 |
#define expect_false(expr) expect ((expr) != 0, 0) |
89 |
#define expect_true(expr) expect ((expr) != 0, 1) |
90 |
|
91 |
/* |
92 |
* compressed format |
93 |
* |
94 |
* 000LLLLL <L+1> ; literal |
95 |
* LLLooooo oooooooo ; backref L |
96 |
* 111ooooo LLLLLLLL oooooooo ; backref L+7 |
97 |
* |
98 |
*/ |
99 |
|
100 |
unsigned int |
101 |
lzf_compress (const void *const in_data, unsigned int in_len, |
102 |
void *out_data, unsigned int out_len |
103 |
#if LZF_STATE_ARG |
104 |
, LZF_STATE htab |
105 |
#endif |
106 |
) |
107 |
{ |
108 |
#if !LZF_STATE_ARG |
109 |
LZF_STATE htab; |
110 |
#endif |
111 |
const u8 **hslot; |
112 |
const u8 *ip = (const u8 *)in_data; |
113 |
u8 *op = (u8 *)out_data; |
114 |
const u8 *in_end = ip + in_len; |
115 |
u8 *out_end = op + out_len; |
116 |
const u8 *ref; |
117 |
|
118 |
unsigned int hval = FRST (ip); |
119 |
unsigned long off; |
120 |
int lit = 0; |
121 |
|
122 |
#if INIT_HTAB |
123 |
memset (htab, 0, sizeof (htab)); |
124 |
# if 0 |
125 |
for (hslot = htab; hslot < htab + HSIZE; hslot++) |
126 |
*hslot++ = ip; |
127 |
# endif |
128 |
#endif |
129 |
|
130 |
for (;;) |
131 |
{ |
132 |
if (expect_true (ip < in_end - 2)) |
133 |
{ |
134 |
hval = NEXT (hval, ip); |
135 |
hslot = htab + IDX (hval); |
136 |
ref = *hslot; *hslot = ip; |
137 |
|
138 |
if (1 |
139 |
#if INIT_HTAB && !USE_MEMCPY |
140 |
&& ref < ip /* the next test will actually take care of this, but this is faster */ |
141 |
#endif |
142 |
&& (off = ip - ref - 1) < MAX_OFF |
143 |
&& ip + 4 < in_end |
144 |
&& ref > (u8 *)in_data |
145 |
#if STRICT_ALIGN |
146 |
x x x |
147 |
&& ref[0] == ip[0] |
148 |
&& ref[1] == ip[1] |
149 |
&& ref[2] == ip[2] |
150 |
#else |
151 |
&& *(u16 *)ref == *(u16 *)ip |
152 |
&& ref[2] == ip[2] |
153 |
#endif |
154 |
) |
155 |
{ |
156 |
/* match found at *ref++ */ |
157 |
unsigned int len = 2; |
158 |
unsigned int maxlen = in_end - ip - len; |
159 |
maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; |
160 |
|
161 |
if (expect_false (op + lit + 1 + 3 >= out_end)) |
162 |
return 0; |
163 |
|
164 |
if (expect_false (lit)) |
165 |
{ |
166 |
*op++ = lit - 1; |
167 |
lit = -lit; |
168 |
do |
169 |
*op++ = ip[lit]; |
170 |
while (expect_false (++lit)); |
171 |
} |
172 |
|
173 |
for (;;) |
174 |
{ |
175 |
if (expect_true (maxlen > 16)) |
176 |
{ |
177 |
len++; if (ref [len] != ip [len]) break; |
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 |
len++; if (ref [len] != ip [len]) break; |
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 |
len++; if (ref [len] != ip [len]) break; |
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 |
len++; if (ref [len] != ip [len]) break; |
193 |
} |
194 |
|
195 |
do |
196 |
len++; |
197 |
while (len < maxlen && ref[len] == ip[len]); |
198 |
|
199 |
break; |
200 |
} |
201 |
|
202 |
len -= 2; |
203 |
ip++; |
204 |
|
205 |
if (len < 7) |
206 |
{ |
207 |
*op++ = (off >> 8) + (len << 5); |
208 |
} |
209 |
else |
210 |
{ |
211 |
*op++ = (off >> 8) + ( 7 << 5); |
212 |
*op++ = len - 7; |
213 |
} |
214 |
|
215 |
*op++ = off; |
216 |
|
217 |
#if ULTRA_FAST || VERY_FAST |
218 |
ip += len; |
219 |
#if VERY_FAST && !ULTRA_FAST |
220 |
--ip; |
221 |
#endif |
222 |
hval = FRST (ip); |
223 |
|
224 |
hval = NEXT (hval, ip); |
225 |
htab[IDX (hval)] = ip; |
226 |
ip++; |
227 |
|
228 |
#if VERY_FAST && !ULTRA_FAST |
229 |
hval = NEXT (hval, ip); |
230 |
htab[IDX (hval)] = ip; |
231 |
ip++; |
232 |
#endif |
233 |
#else |
234 |
do |
235 |
{ |
236 |
hval = NEXT (hval, ip); |
237 |
htab[IDX (hval)] = ip; |
238 |
ip++; |
239 |
} |
240 |
while (len--); |
241 |
#endif |
242 |
continue; |
243 |
} |
244 |
} |
245 |
else if (expect_false (ip == in_end)) |
246 |
break; |
247 |
|
248 |
/* one more literal byte we must copy */ |
249 |
lit++; |
250 |
ip++; |
251 |
|
252 |
if (expect_false (lit == MAX_LIT)) |
253 |
{ |
254 |
if (op + 1 + MAX_LIT >= out_end) |
255 |
return 0; |
256 |
|
257 |
*op++ = MAX_LIT - 1; |
258 |
|
259 |
#ifdef lzf_movsb |
260 |
ip -= MAX_LIT; |
261 |
lzf_movsb (op, ip, lit); |
262 |
#else |
263 |
lit = -lit; |
264 |
do |
265 |
*op++ = ip[lit]; |
266 |
while (++lit); |
267 |
#endif |
268 |
} |
269 |
} |
270 |
|
271 |
if (lit) |
272 |
{ |
273 |
if (op + lit + 1 >= out_end) |
274 |
return 0; |
275 |
|
276 |
*op++ = lit - 1; |
277 |
#ifdef lzf_movsb |
278 |
ip -= lit; |
279 |
lzf_movsb (op, ip, lit); |
280 |
#else |
281 |
lit = -lit; |
282 |
do |
283 |
*op++ = ip[lit]; |
284 |
while (++lit); |
285 |
#endif |
286 |
} |
287 |
|
288 |
return op - (u8 *) out_data; |
289 |
} |
290 |
|