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 1.5 by pcg, Mon Dec 29 12:48:16 2003 UTC vs.
Revision 1.40 by root, Tue Jun 1 01:50:24 2010 UTC

1/* 1/*
2 * Copyright (c) 2000-2003 Marc Alexander Lehmann <pcg@goof.com> 2 * Copyright (c) 2000-2010 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-
22 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 19 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
23 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 20 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
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.
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.
27 */ 35 */
28 36
29#include "lzfP.h" 37#include "lzfP.h"
30 38
31#define HSIZE (1 << (HLOG)) 39#define HSIZE (1 << (HLOG))
34 * don't play with this unless you benchmark! 42 * don't play with this unless you benchmark!
35 * decompression is not dependent on the hash function 43 * decompression is not dependent on the hash function
36 * the hashing function might seem strange, just believe me 44 * the hashing function might seem strange, just believe me
37 * it works ;) 45 * it works ;)
38 */ 46 */
47#ifndef FRST
39#define FRST(p) (((p[0]) << 8) + p[1]) 48# define FRST(p) (((p[0]) << 8) | p[1])
40#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
41#define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) + h*3) & (HSIZE - 1)) 55# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
56# endif
57#endif
42/* 58/*
43 * 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.
44 * (h * 57321 >> (3*8 - HLOG)) 60 * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1))
61 * the latter is also quite fast on newer CPUs, and compresses similarly.
62 *
45 * the next one is also quite good, albeit slow ;) 63 * the next one is also quite good, albeit slow ;)
46 * (int)(cos(h & 0xffffff) * 1e6) 64 * (int)(cos(h & 0xffffff) * 1e6)
47 */ 65 */
48 66
49#if 0 67#if 0
50/* original lzv-like hash function */ 68/* original lzv-like hash function, much worse and thus slower */
51# define FRST(p) (p[0] << 5) ^ p[1] 69# define FRST(p) (p[0] << 5) ^ p[1]
52# define NEXT(v,p) ((v) << 5) ^ p[2] 70# define NEXT(v,p) ((v) << 5) ^ p[2]
53# define IDX(h) ((h) & (HSIZE - 1)) 71# define IDX(h) ((h) & (HSIZE - 1))
54#endif 72#endif
55 73
56#define MAX_LIT (1 << 5) 74#define MAX_LIT (1 << 5)
57#define MAX_OFF (1 << 13) 75#define MAX_OFF (1 << 13)
58#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)
59 88
60/* 89/*
61 * compressed format 90 * compressed format
62 * 91 *
63 * 000LLLLL <L+1> ; literal 92 * 000LLLLL <L+1> ; literal, L+1=1..33 octets
64 * LLLOOOOO oooooooo ; backref L 93 * LLLooooo oooooooo ; backref L+1=1..7 octets, o+1=1..4096 offset
65 * 111OOOOO LLLLLLLL oooooooo ; backref L+7 94 * 111ooooo LLLLLLLL oooooooo ; backref L+8 octets, o+1=1..4096 offset
66 * 95 *
67 */ 96 */
68 97
69unsigned int 98unsigned int
70lzf_compress (const void *const in_data, unsigned int in_len, 99lzf_compress (const void *const in_data, unsigned int in_len,
71 void *out_data, unsigned int out_len 100 void *out_data, unsigned int out_len
72#if LZF_STATE_ARG 101#if LZF_STATE_ARG
73 , LZF_STATE *htab 102 , LZF_STATE htab
74#endif 103#endif
75 ) 104 )
76{ 105{
77#if !LZF_STATE_ARG 106#if !LZF_STATE_ARG
78 LZF_STATE htab; 107 LZF_STATE htab;
82 u8 *op = (u8 *)out_data; 111 u8 *op = (u8 *)out_data;
83 const u8 *in_end = ip + in_len; 112 const u8 *in_end = ip + in_len;
84 u8 *out_end = op + out_len; 113 u8 *out_end = op + out_len;
85 const u8 *ref; 114 const u8 *ref;
86 115
87 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
88 unsigned long off; 126 unsigned long off;
89 int lit = 0; 127#endif
128 unsigned int hval;
129 int lit;
130
131 if (!in_len || !out_len)
132 return 0;
90 133
91#if INIT_HTAB 134#if INIT_HTAB
92# if USE_MEMCPY
93 memset (htab, 0, sizeof (htab)); 135 memset (htab, 0, sizeof (htab));
94# else 136# if 0
95 for (hslot = htab; hslot < htab + HSIZE; hslot++) 137 for (hslot = htab; hslot < htab + HSIZE; hslot++)
96 *hslot++ = ip; 138 *hslot++ = ip;
97# endif
98#endif 139# endif
140#endif
99 141
100 for (;;) 142 lit = 0; op++; /* start run */
143
144 hval = FRST (ip);
145 while (ip < in_end - 2)
101 { 146 {
102 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 && ref[2] == ip[2]
159#if STRICT_ALIGN
160 && (ref[1] << 8) | ref[0] == (ip[1] << 8) | ip[0]
161#else
162 && *(u16 *)ref == *(u16 *)ip
163#endif
164 )
103 { 165 {
104 hval = NEXT (hval, ip);
105 hslot = htab + IDX (hval);
106 ref = *hslot; *hslot = ip;
107
108 if (1
109#if INIT_HTAB && !USE_MEMCPY
110 && ref < ip /* the next test will actually take care of this, but this is faster */
111#endif
112 && (off = ip - ref - 1) < MAX_OFF
113 && ip + 4 < in_end
114 && ref > (u8 *)in_data
115#if STRICT_ALIGN
116 && ref[0] == ip[0]
117 && ref[1] == ip[1]
118 && ref[2] == ip[2]
119#else
120 && *(u16 *)ref == *(u16 *)ip
121 && ref[2] == ip[2]
122#endif
123 )
124 {
125 /* match found at *ref++ */ 166 /* match found at *ref++ */
126 unsigned int len = 2; 167 unsigned int len = 2;
127 unsigned int maxlen = in_end - ip - len; 168 unsigned int maxlen = in_end - ip - len;
128 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; 169 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen;
170
171 if (expect_false (op + 3 + 1 >= out_end)) /* first a faster conservative test */
172 if (op - !lit + 3 + 1 >= out_end) /* second the exact but rare test */
173 return 0;
174
175 op [- lit - 1] = lit - 1; /* stop run */
176 op -= !lit; /* undo run if length is zero */
177
178 for (;;)
179 {
180 if (expect_true (maxlen > 16))
181 {
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
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
192 len++; if (ref [len] != ip [len]) break;
193 len++; if (ref [len] != ip [len]) break;
194 len++; if (ref [len] != ip [len]) break;
195 len++; if (ref [len] != ip [len]) break;
196
197 len++; if (ref [len] != ip [len]) break;
198 len++; if (ref [len] != ip [len]) break;
199 len++; if (ref [len] != ip [len]) break;
200 len++; if (ref [len] != ip [len]) break;
201 }
129 202
130 do 203 do
131 len++; 204 len++;
132 while (len < maxlen && ref[len] == ip[len]); 205 while (len < maxlen && ref[len] == ip[len]);
133 206
134 if (op + lit + 1 + 3 >= out_end)
135 return 0;
136
137 if (lit)
138 { 207 break;
139 *op++ = lit - 1;
140 lit = -lit;
141 do
142 *op++ = ip[lit];
143 while (++lit);
144 } 208 }
145 209
146 len -= 2; 210 len -= 2; /* len is now #octets - 1 */
147 ip++; 211 ip++;
148 212
149 if (len < 7) 213 if (len < 7)
150 { 214 {
151 *op++ = (off >> 8) + (len << 5); 215 *op++ = (off >> 8) + (len << 5);
152 } 216 }
153 else 217 else
154 { 218 {
155 *op++ = (off >> 8) + ( 7 << 5); 219 *op++ = (off >> 8) + ( 7 << 5);
156 *op++ = len - 7; 220 *op++ = len - 7;
157 } 221 }
158 222
159 *op++ = off; 223 *op++ = off;
224 lit = 0; op++; /* start run */
160 225
161#if ULTRA_FAST
162 ip += len; 226 ip += len + 1;
227
228 if (expect_false (ip >= in_end - 2))
229 break;
230
231#if ULTRA_FAST || VERY_FAST
232 --ip;
233# if VERY_FAST && !ULTRA_FAST
234 --ip;
235# endif
163 hval = FRST (ip); 236 hval = FRST (ip);
237
238 hval = NEXT (hval, ip);
239 htab[IDX (hval)] = ip;
240 ip++;
241
242# if VERY_FAST && !ULTRA_FAST
243 hval = NEXT (hval, ip);
244 htab[IDX (hval)] = ip;
245 ip++;
246# endif
247#else
248 ip -= len + 1;
249
250 do
251 {
164 hval = NEXT (hval, ip); 252 hval = NEXT (hval, ip);
165 htab[IDX (hval)] = ip; 253 htab[IDX (hval)] = ip;
166 ip++; 254 ip++;
167#else
168 do
169 {
170 hval = NEXT (hval, ip);
171 htab[IDX (hval)] = ip;
172 ip++;
173 } 255 }
174 while (len--); 256 while (len--);
175#endif 257#endif
176 continue;
177 }
178 } 258 }
179 else if (ip == in_end) 259 else
180 break;
181
182 /* one more literal byte we must copy */
183 lit++;
184 ip++;
185
186 if (lit == MAX_LIT)
187 { 260 {
188 if (op + 1 + MAX_LIT >= out_end) 261 /* one more literal byte we must copy */
262 if (expect_false (op >= out_end))
189 return 0; 263 return 0;
190 264
191 *op++ = MAX_LIT - 1; 265 lit++; *op++ = *ip++;
192#if USE_MEMCPY 266
193 memcpy (op, ip - MAX_LIT, MAX_LIT); 267 if (expect_false (lit == MAX_LIT))
194 op += MAX_LIT;
195 lit = 0;
196#else
197 lit = -lit;
198 do 268 {
199 *op++ = ip[lit]; 269 op [- lit - 1] = lit - 1; /* stop run */
200 while (++lit); 270 lit = 0; op++; /* start run */
201#endif 271 }
202 } 272 }
203 } 273 }
204 274
205 if (lit) 275 if (op + 3 > out_end) /* at most 3 bytes can be missing here */
276 return 0;
277
278 while (ip < in_end)
206 { 279 {
207 if (op + lit + 1 >= out_end) 280 lit++; *op++ = *ip++;
208 return 0;
209 281
210 *op++ = lit - 1; 282 if (expect_false (lit == MAX_LIT))
211 lit = -lit; 283 {
212 do 284 op [- lit - 1] = lit - 1; /* stop run */
213 *op++ = ip[lit]; 285 lit = 0; op++; /* start run */
214 while (++lit); 286 }
215 } 287 }
216 288
289 op [- lit - 1] = lit - 1; /* end run */
290 op -= !lit; /* undo run if length is zero */
291
217 return op - (u8 *) out_data; 292 return op - (u8 *)out_data;
218} 293}
294

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines