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 1.45 by root, Sun Mar 27 23:53:23 2011 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))
32 40
33/* 41/*
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 * the data format is not dependent on the hash function.
36 * the hashing function might seem strange, just believe me 44 * the hash 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 MULTIPLICATION_IS_SLOW
51# if ULTRA_FAST
52# define IDX(h) ((( h >> (3*8 - HLOG)) - h ) & (HSIZE - 1))
53# elif VERY_FAST
54# define IDX(h) ((( h >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
55# else
41#define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) + h*3) & (HSIZE - 1)) 56# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
42/* 57# endif
43 * IDX works because it is very similar to a multiplicative hash, e.g. 58# else
44 * (h * 57321 >> (3*8 - HLOG)) 59/* this one was developed with sesse,
45 * the next one is also quite good, albeit slow ;) 60 * and is very similar to the one in snappy.
46 * (int)(cos(h & 0xffffff) * 1e6) 61 * it does need a modern enough cpu with a fast multiplication.
47 */ 62 */
63# define IDX(h) (((h * 0x1e35a7bdU) >> (32 - HLOG - 8)) & (HSIZE - 1))
64# endif
65#endif
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;
79#endif 108#endif
80 const u8 **hslot;
81 const u8 *ip = (const u8 *)in_data; 109 const u8 *ip = (const u8 *)in_data;
82 u8 *op = (u8 *)out_data; 110 u8 *op = (u8 *)out_data;
83 const u8 *in_end = ip + in_len; 111 const u8 *in_end = ip + in_len;
84 u8 *out_end = op + out_len; 112 u8 *out_end = op + out_len;
85 const u8 *ref; 113 const u8 *ref;
86 114
87 unsigned int hval = FRST (ip); 115 /* off requires a type wide enough to hold a general pointer difference.
116 * ISO C doesn't have that (size_t might not be enough and ptrdiff_t only
117 * works for differences within a single object). We also assume that no
118 * no bit pattern traps. Since the only platform that is both non-POSIX
119 * and fails to support both assumptions is windows 64 bit, we make a
120 * special workaround for it.
121 */
122#if defined (WIN32) && defined (_M_X64)
123 unsigned _int64 off; /* workaround for missing POSIX compliance */
124#else
88 unsigned long off; 125 unsigned long off;
89 int lit = 0; 126#endif
127 unsigned int hval;
128 int lit;
129
130 if (!in_len || !out_len)
131 return 0;
90 132
91#if INIT_HTAB 133#if INIT_HTAB
92# if USE_MEMCPY
93 memset (htab, 0, sizeof (htab)); 134 memset (htab, 0, sizeof (htab));
94# else
95 for (hslot = htab; hslot < htab + HSIZE; hslot++)
96 *hslot++ = ip;
97# endif 135#endif
98#endif
99 136
100 for (;;) 137 lit = 0; op++; /* start run */
138
139 hval = FRST (ip);
140 while (ip < in_end - 2)
101 { 141 {
102 if (ip < in_end - 2) 142 LZF_HSLOT *hslot;
143
144 hval = NEXT (hval, ip);
145 hslot = htab + IDX (hval);
146 ref = *hslot + LZF_HSLOT_BIAS; *hslot = ip - LZF_HSLOT_BIAS;
147
148 if (1
149#if INIT_HTAB
150 && ref < ip /* the next test will actually take care of this, but this is faster */
151#endif
152 && (off = ip - ref - 1) < MAX_OFF
153 && ref > (u8 *)in_data
154 && ref[2] == ip[2]
155#if STRICT_ALIGN
156 && ((ref[1] << 8) | ref[0]) == ((ip[1] << 8) | ip[0])
157#else
158 && *(u16 *)ref == *(u16 *)ip
159#endif
160 )
103 { 161 {
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++ */ 162 /* match found at *ref++ */
126 unsigned int len = 2; 163 unsigned int len = 2;
127 unsigned int maxlen = in_end - ip - len; 164 unsigned int maxlen = in_end - ip - len;
128 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; 165 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen;
166
167 if (expect_false (op + 3 + 1 >= out_end)) /* first a faster conservative test */
168 if (op - !lit + 3 + 1 >= out_end) /* second the exact but rare test */
169 return 0;
170
171 op [- lit - 1] = lit - 1; /* stop run */
172 op -= !lit; /* undo run if length is zero */
173
174 for (;;)
175 {
176 if (expect_true (maxlen > 16))
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 len++; if (ref [len] != ip [len]) break;
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 }
129 198
130 do 199 do
131 len++; 200 len++;
132 while (len < maxlen && ref[len] == ip[len]); 201 while (len < maxlen && ref[len] == ip[len]);
133 202
134 if (op + lit + 1 + 3 >= out_end)
135 return 0;
136
137 if (lit)
138 { 203 break;
139 *op++ = lit - 1;
140 lit = -lit;
141 do
142 *op++ = ip[lit];
143 while (++lit);
144 } 204 }
145 205
206 len -= 2; /* len is now #octets - 1 */
207 ip++;
208
209 if (len < 7)
210 {
211 *op++ = (off >> 8) + (len << 5);
212 }
213 else
214 {
215 *op++ = (off >> 8) + ( 7 << 5);
146 len -= 2; 216 *op++ = len - 7;
217 }
218
219 *op++ = off;
220
221 lit = 0; op++; /* start run */
222
223 ip += len + 1;
224
225 if (expect_false (ip >= in_end - 2))
226 break;
227
228#if ULTRA_FAST || VERY_FAST
229 --ip;
230# if VERY_FAST && !ULTRA_FAST
231 --ip;
232# endif
233 hval = FRST (ip);
234
235 hval = NEXT (hval, ip);
236 htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
237 ip++;
238
239# if VERY_FAST && !ULTRA_FAST
240 hval = NEXT (hval, ip);
241 htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
242 ip++;
243# endif
244#else
245 ip -= len + 1;
246
247 do
248 {
249 hval = NEXT (hval, ip);
250 htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
147 ip++; 251 ip++;
148
149 if (len < 7)
150 {
151 *op++ = (off >> 8) + (len << 5);
152 } 252 }
153 else
154 {
155 *op++ = (off >> 8) + ( 7 << 5);
156 *op++ = len - 7;
157 }
158
159 *op++ = off;
160
161#if ULTRA_FAST
162 ip += len;
163 hval = FRST (ip);
164 hval = NEXT (hval, ip);
165 htab[IDX (hval)] = ip;
166 ip++;
167#else
168 do
169 {
170 hval = NEXT (hval, ip);
171 htab[IDX (hval)] = ip;
172 ip++;
173 }
174 while (len--); 253 while (len--);
175#endif 254#endif
176 continue;
177 }
178 } 255 }
179 else if (ip == in_end) 256 else
180 break;
181
182 /* one more literal byte we must copy */
183 lit++;
184 ip++;
185
186 if (lit == MAX_LIT)
187 { 257 {
188 if (op + 1 + MAX_LIT >= out_end) 258 /* one more literal byte we must copy */
259 if (expect_false (op >= out_end))
189 return 0; 260 return 0;
190 261
191 *op++ = MAX_LIT - 1; 262 lit++; *op++ = *ip++;
192#if USE_MEMCPY 263
193 memcpy (op, ip - MAX_LIT, MAX_LIT); 264 if (expect_false (lit == MAX_LIT))
194 op += MAX_LIT;
195 lit = 0;
196#else
197 lit = -lit;
198 do 265 {
199 *op++ = ip[lit]; 266 op [- lit - 1] = lit - 1; /* stop run */
200 while (++lit); 267 lit = 0; op++; /* start run */
201#endif 268 }
202 } 269 }
203 } 270 }
204 271
205 if (lit) 272 if (op + 3 > out_end) /* at most 3 bytes can be missing here */
273 return 0;
274
275 while (ip < in_end)
206 { 276 {
207 if (op + lit + 1 >= out_end) 277 lit++; *op++ = *ip++;
208 return 0;
209 278
210 *op++ = lit - 1; 279 if (expect_false (lit == MAX_LIT))
211 lit = -lit; 280 {
212 do 281 op [- lit - 1] = lit - 1; /* stop run */
213 *op++ = ip[lit]; 282 lit = 0; op++; /* start run */
214 while (++lit); 283 }
215 } 284 }
216 285
286 op [- lit - 1] = lit - 1; /* end run */
287 op -= !lit; /* undo run if length is zero */
288
217 return op - (u8 *) out_data; 289 return op - (u8 *)out_data;
218} 290}
291

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines