ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/liblzf/lzf_c.c
(Generate patch)

Comparing liblzf/lzf_c.c (file contents):
Revision 1.8 by root, Thu Mar 3 17:06:44 2005 UTC vs.
Revision 1.38 by root, Sun May 30 05:46:56 2010 UTC

1/* 1/*
2 * Copyright (c) 2000-2005 Marc Alexander Lehmann <schmorp@schmorp.de> 2 * Copyright (c) 2000-2009 Marc Alexander Lehmann <schmorp@schmorp.de>
3 * 3 *
4 * Redistribution and use in source and binary forms, with or without modifica- 4 * Redistribution and use in source and binary forms, with or without modifica-
5 * tion, are permitted provided that the following conditions are met: 5 * tion, are permitted provided that the following conditions are met:
6 * 6 *
7 * 1. Redistributions of source code must retain the above copyright notice, 7 * 1. Redistributions of source code must retain the above copyright notice,
8 * this list of conditions and the following disclaimer. 8 * this list of conditions and the following disclaimer.
9 * 9 *
10 * 2. Redistributions in binary form must reproduce the above copyright 10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the 11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution. 12 * documentation and/or other materials provided with the distribution.
13 *
14 * 3. The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 * 13 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
18 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER- 15 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
19 * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO 16 * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
20 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE- 17 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
24 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- 21 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
25 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED 22 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
26 * OF THE POSSIBILITY OF SUCH DAMAGE. 23 * OF THE POSSIBILITY OF SUCH DAMAGE.
27 * 24 *
28 * Alternatively, the contents of this file may be used under the terms of 25 * Alternatively, the contents of this file may be used under the terms of
29 * the GNU General Public License version 2 (the "GPL"), in which case the 26 * the GNU General Public License ("GPL") version 2 or any later version,
30 * provisions of the GPL are applicable instead of the above. If you wish to 27 * in which case the provisions of the GPL are applicable instead of
31 * allow the use of your version of this file only under the terms of the 28 * the above. If you wish to allow the use of your version of this file
32 * GPL and not to allow others to use your version of this file under the 29 * only under the terms of the GPL and not to allow others to use your
33 * BSD license, indicate your decision by deleting the provisions above and 30 * version of this file under the BSD license, indicate your decision
34 * replace them with the notice and other provisions required by the GPL. If 31 * by deleting the provisions above and replace them with the notice
35 * you do not delete the provisions above, a recipient may use your version 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
36 * of this file under either the BSD or the GPL. 34 * either the BSD or the GPL.
37 */ 35 */
38 36
39#include "lzfP.h" 37#include "lzfP.h"
40 38
41#define HSIZE (1 << (HLOG)) 39#define HSIZE (1 << (HLOG))
44 * don't play with this unless you benchmark! 42 * don't play with this unless you benchmark!
45 * decompression is not dependent on the hash function 43 * decompression is not dependent on the hash function
46 * the hashing function might seem strange, just believe me 44 * the hashing function might seem strange, just believe me
47 * it works ;) 45 * it works ;)
48 */ 46 */
47#ifndef FRST
49#define FRST(p) (((p[0]) << 8) + p[1]) 48# define FRST(p) (((p[0]) << 8) | p[1])
50#define NEXT(v,p) (((v) << 8) + p[2]) 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
51#define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1)) 55# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
56# endif
57#endif
52/* 58/*
53 * IDX works because it is very similar to a multiplicative hash, e.g. 59 * IDX works because it is very similar to a multiplicative hash, e.g.
54 * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1)) 60 * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1))
55 * the latter is also quite fast on newer CPUs, and sligthly better 61 * the latter is also quite fast on newer CPUs, and compresses similarly.
56 * 62 *
57 * the next one is also quite good, albeit slow ;) 63 * the next one is also quite good, albeit slow ;)
58 * (int)(cos(h & 0xffffff) * 1e6) 64 * (int)(cos(h & 0xffffff) * 1e6)
59 */ 65 */
60 66
65# define IDX(h) ((h) & (HSIZE - 1)) 71# define IDX(h) ((h) & (HSIZE - 1))
66#endif 72#endif
67 73
68#define MAX_LIT (1 << 5) 74#define MAX_LIT (1 << 5)
69#define MAX_OFF (1 << 13) 75#define MAX_OFF (1 << 13)
70#define MAX_REF ((1 << 8) + (1 << 3)) 76#define MAX_REF ((1 << 8) + (1 << 3))
77
78#if __GNUC__ >= 3
79# define expect(expr,value) __builtin_expect ((expr),(value))
80# define inline inline
81#else
82# define expect(expr,value) (expr)
83# define inline static
84#endif
85
86#define expect_false(expr) expect ((expr) != 0, 0)
87#define expect_true(expr) expect ((expr) != 0, 1)
71 88
72/* 89/*
73 * compressed format 90 * compressed format
74 * 91 *
75 * 000LLLLL <L+1> ; literal 92 * 000LLLLL <L+1> ; literal, L+1=1..33 octets
76 * LLLOOOOO oooooooo ; backref L 93 * LLLooooo oooooooo ; backref L+1=1..7 octets, o+1=1..4096 offset
77 * 111OOOOO LLLLLLLL oooooooo ; backref L+7 94 * 111ooooo LLLLLLLL oooooooo ; backref L+8 octets, o+1=1..4096 offset
78 * 95 *
79 */ 96 */
80 97
81unsigned int 98unsigned int
82lzf_compress (const void *const in_data, unsigned int in_len, 99lzf_compress (const void *const in_data, unsigned int in_len,
83 void *out_data, unsigned int out_len 100 void *out_data, unsigned int out_len
84#if LZF_STATE_ARG 101#if LZF_STATE_ARG
85 , LZF_STATE *htab 102 , LZF_STATE htab
86#endif 103#endif
87 ) 104 )
88{ 105{
89#if !LZF_STATE_ARG 106#if !LZF_STATE_ARG
90 LZF_STATE htab; 107 LZF_STATE htab;
94 u8 *op = (u8 *)out_data; 111 u8 *op = (u8 *)out_data;
95 const u8 *in_end = ip + in_len; 112 const u8 *in_end = ip + in_len;
96 u8 *out_end = op + out_len; 113 u8 *out_end = op + out_len;
97 const u8 *ref; 114 const u8 *ref;
98 115
99 unsigned int hval = FRST (ip); 116 /* off requires a type wide enough to hold a general pointer difference.
117 * ISO C doesn't have that (size_t might not be enough and ptrdiff_t only
118 * works for differences within a single object). We also assume that no
119 * no bit pattern traps. Since the only platform that is both non-POSIX
120 * and fails to support both assumptions is windows 64 bit, we make a
121 * special workaround for it.
122 */
123#if defined (WIN32) && defined (_M_X64)
124 unsigned _int64 off; /* workaround for missing POSIX compliance */
125#else
100 unsigned long off; 126 unsigned long off;
101 int lit = 0; 127#endif
128 unsigned int hval;
129 int lit;
130
131 if (!in_len || !out_len)
132 return 0;
102 133
103#if INIT_HTAB 134#if INIT_HTAB
104# if USE_MEMCPY
105 memset (htab, 0, sizeof (htab)); 135 memset (htab, 0, sizeof (htab));
106# else 136# if 0
107 for (hslot = htab; hslot < htab + HSIZE; hslot++) 137 for (hslot = htab; hslot < htab + HSIZE; hslot++)
108 *hslot++ = ip; 138 *hslot++ = ip;
109# endif
110#endif 139# endif
140#endif
111 141
112 for (;;) 142 lit = 0; op++; /* start run */
143
144 hval = FRST (ip);
145 while (ip < in_end - 2)
113 { 146 {
114 if (ip < in_end - 2) 147 hval = NEXT (hval, ip);
148 hslot = htab + IDX (hval);
149 ref = *hslot; *hslot = ip;
150
151 if (1
152#if INIT_HTAB
153 && ref < ip /* the next test will actually take care of this, but this is faster */
154#endif
155 && (off = ip - ref - 1) < MAX_OFF
156 && ip + 4 < in_end
157 && ref > (u8 *)in_data
158#if STRICT_ALIGN
159 && ref[0] == ip[0]
160 && ref[1] == ip[1]
161 && ref[2] == ip[2]
162#else
163 && *(u16 *)ref == *(u16 *)ip
164 && ref[2] == ip[2]
165#endif
166 )
115 { 167 {
116 hval = NEXT (hval, ip);
117 hslot = htab + IDX (hval);
118 ref = *hslot; *hslot = ip;
119
120 if (1
121#if INIT_HTAB && !USE_MEMCPY
122 && ref < ip /* the next test will actually take care of this, but this is faster */
123#endif
124 && (off = ip - ref - 1) < MAX_OFF
125 && ip + 4 < in_end
126 && ref > (u8 *)in_data
127#if STRICT_ALIGN
128 && ref[0] == ip[0]
129 && ref[1] == ip[1]
130 && ref[2] == ip[2]
131#else
132 && *(u16 *)ref == *(u16 *)ip
133 && ref[2] == ip[2]
134#endif
135 )
136 {
137 /* match found at *ref++ */ 168 /* match found at *ref++ */
138 unsigned int len = 2; 169 unsigned int len = 2;
139 unsigned int maxlen = in_end - ip - len; 170 unsigned int maxlen = in_end - ip - len;
140 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; 171 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen;
172
173 if (expect_false (op + 3 + 1 >= out_end)) /* first a faster conservative test */
174 if (op - !lit + 3 + 1 >= out_end) /* second the exact but rare test */
175 return 0;
176
177 op [- lit - 1] = lit - 1; /* stop run */
178 op -= !lit; /* undo run if length is zero */
179
180 for (;;)
181 {
182 if (expect_true (maxlen > 16))
183 {
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
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 len++; if (ref [len] != ip [len]) break;
195 len++; if (ref [len] != ip [len]) break;
196 len++; if (ref [len] != ip [len]) break;
197 len++; if (ref [len] != ip [len]) break;
198
199 len++; if (ref [len] != ip [len]) break;
200 len++; if (ref [len] != ip [len]) break;
201 len++; if (ref [len] != ip [len]) break;
202 len++; if (ref [len] != ip [len]) break;
203 }
141 204
142 do 205 do
143 len++; 206 len++;
144 while (len < maxlen && ref[len] == ip[len]); 207 while (len < maxlen && ref[len] == ip[len]);
145 208
146 if (op + lit + 1 + 3 >= out_end)
147 return 0;
148
149 if (lit)
150 { 209 break;
151 *op++ = lit - 1;
152 lit = -lit;
153 do
154 *op++ = ip[lit];
155 while (++lit);
156 } 210 }
157 211
158 len -= 2; 212 len -= 2; /* len is now #octets - 1 */
159 ip++; 213 ip++;
160 214
161 if (len < 7) 215 if (len < 7)
162 { 216 {
163 *op++ = (off >> 8) + (len << 5); 217 *op++ = (off >> 8) + (len << 5);
164 } 218 }
165 else 219 else
166 { 220 {
167 *op++ = (off >> 8) + ( 7 << 5); 221 *op++ = (off >> 8) + ( 7 << 5);
168 *op++ = len - 7; 222 *op++ = len - 7;
169 } 223 }
170 224
171 *op++ = off; 225 *op++ = off;
226 lit = 0; op++; /* start run */
172 227
173#if ULTRA_FAST
174 ip += len; 228 ip += len + 1;
229
230 if (expect_false (ip >= in_end - 2))
231 break;
232
233#if ULTRA_FAST || VERY_FAST
234 --ip;
235# if VERY_FAST && !ULTRA_FAST
236 --ip;
237# endif
175 hval = FRST (ip); 238 hval = FRST (ip);
239
240 hval = NEXT (hval, ip);
241 htab[IDX (hval)] = ip;
242 ip++;
243
244# if VERY_FAST && !ULTRA_FAST
245 hval = NEXT (hval, ip);
246 htab[IDX (hval)] = ip;
247 ip++;
248# endif
249#else
250 ip -= len + 1;
251
252 do
253 {
176 hval = NEXT (hval, ip); 254 hval = NEXT (hval, ip);
177 htab[IDX (hval)] = ip; 255 htab[IDX (hval)] = ip;
178 ip++; 256 ip++;
179#else
180 do
181 {
182 hval = NEXT (hval, ip);
183 htab[IDX (hval)] = ip;
184 ip++;
185 } 257 }
186 while (len--); 258 while (len--);
187#endif 259#endif
188 continue;
189 }
190 } 260 }
191 else if (ip == in_end) 261 else
192 break;
193
194 /* one more literal byte we must copy */
195 lit++;
196 ip++;
197
198 if (lit == MAX_LIT)
199 { 262 {
200 if (op + 1 + MAX_LIT >= out_end) 263 /* one more literal byte we must copy */
264 if (expect_false (op >= out_end))
201 return 0; 265 return 0;
202 266
203 *op++ = MAX_LIT - 1; 267 lit++; *op++ = *ip++;
204#if USE_MEMCPY 268
205 memcpy (op, ip - MAX_LIT, MAX_LIT); 269 if (expect_false (lit == MAX_LIT))
206 op += MAX_LIT;
207 lit = 0;
208#else
209 lit = -lit;
210 do 270 {
211 *op++ = ip[lit]; 271 op [- lit - 1] = lit - 1; /* stop run */
212 while (++lit); 272 lit = 0; op++; /* start run */
213#endif 273 }
214 } 274 }
215 } 275 }
216 276
217 if (lit) 277 if (op + 3 > out_end) /* at most 3 bytes can be missing here */
278 return 0;
279
280 while (ip < in_end)
218 { 281 {
219 if (op + lit + 1 >= out_end) 282 lit++; *op++ = *ip++;
220 return 0;
221 283
222 *op++ = lit - 1; 284 if (expect_false (lit == MAX_LIT))
223 lit = -lit; 285 {
224 do 286 op [- lit - 1] = lit - 1; /* stop run */
225 *op++ = ip[lit]; 287 lit = 0; op++; /* start run */
226 while (++lit); 288 }
227 } 289 }
228 290
291 op [- lit - 1] = lit - 1; /* end run */
292 op -= !lit; /* undo run if length is zero */
293
229 return op - (u8 *) out_data; 294 return op - (u8 *)out_data;
230} 295}
296

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines