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 1.1 by root, Sun Jun 9 22:41:34 2002 UTC vs.
Revision 1.18 by root, Tue Nov 13 08:17:38 2007 UTC

1/* 1/*
2 * Copyright (c) 2000 Marc Alexander Lehmann <pcg@goof.com> 2 * Copyright (c) 2000-2007 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])
41#define IDX(h) ((((h ^ (h << 4)) >> (3*8 - HLOG)) + h*3) & (HSIZE - 1)) 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
42/* 53/*
43 * IDX works because it is very similar to a multiplicative hash, e.g. 54 * IDX works because it is very similar to a multiplicative hash, e.g.
44 * (h * 57321 >> (3*8 - HLOG)) 55 * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1))
56 * the latter is also quite fast on newer CPUs, and compresses similarly.
57 *
45 * the next one is also quite good, albeit slow ;) 58 * the next one is also quite good, albeit slow ;)
46 * (int)(cos(h & 0xffffff) * 1e6) 59 * (int)(cos(h & 0xffffff) * 1e6)
47 */ 60 */
48 61
49#if 0 62#if 0
50/* original lzv-like hash function */ 63/* original lzv-like hash function, much worse and thus slower */
51# define FRST(p) (p[0] << 5) ^ p[1] 64# define FRST(p) (p[0] << 5) ^ p[1]
52# define NEXT(v,p) (v << 5) ^ p[2] 65# define NEXT(v,p) ((v) << 5) ^ p[2]
53# define IDX(h) (h) & (HSIZE - 1) 66# define IDX(h) ((h) & (HSIZE - 1))
54#endif 67#endif
55 68
56#define MAX_LIT (1 << 5) 69#define MAX_LIT (1 << 5)
57#define MAX_OFF (1 << 13) 70#define MAX_OFF (1 << 13)
58#define MAX_REF ((1 << 8) + (1 << 3)) 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)
59 90
60/* 91/*
61 * compressed format 92 * compressed format
62 * 93 *
63 * 000LLLLL <L+1> ; literal 94 * 000LLLLL <L+1> ; literal
64 * LLLOOOOO oooooooo ; backref L 95 * LLLooooo oooooooo ; backref L
65 * 111OOOOO LLLLLLLL oooooooo ; backref L+7 96 * 111ooooo LLLLLLLL oooooooo ; backref L+7
66 * 97 *
67 */ 98 */
68 99
69unsigned int 100unsigned int
70lzf_compress (const void *const in_data, unsigned int in_len, 101lzf_compress (const void *const in_data, unsigned int in_len,
71 void *out_data, unsigned int out_len) 102 void *out_data, unsigned int out_len
103#if LZF_STATE_ARG
104 , LZF_STATE htab
105#endif
106 )
72{ 107{
73 const u8 *htab[HSIZE]; 108#if !LZF_STATE_ARG
109 LZF_STATE htab;
110#endif
74 const u8 **hslot; 111 const u8 **hslot;
75 const u8 *ip = in_data; 112 const u8 *ip = (const u8 *)in_data;
76 u8 *op = out_data; 113 u8 *op = (u8 *)out_data;
77 const u8 *in_end = ip + in_len; 114 const u8 *in_end = ip + in_len;
78 u8 *out_end = op + out_len; 115 u8 *out_end = op + out_len;
79 const u8 *ref; 116 const u8 *ref;
80 117
81 unsigned int hval = FRST (ip); 118 unsigned int hval = FRST (ip);
82 unsigned long off; 119 unsigned long off;
83 int lit = 0; 120 int lit = 0;
84 121
85#if INIT_HTAB 122#if INIT_HTAB
86# if USE_MEMCPY
87 memset (htab, 0, sizeof (htab)); 123 memset (htab, 0, sizeof (htab));
88# else 124# if 0
89 for (hslot = htab; hslot < htab + HSIZE; hslot++) 125 for (hslot = htab; hslot < htab + HSIZE; hslot++)
90 *hslot++ = ip; 126 *hslot++ = ip;
91# endif 127# endif
92#endif 128#endif
93 129
94 do 130 for (;;)
95 { 131 {
132 if (expect_true (ip < in_end - 2))
133 {
96 hval = NEXT (hval, ip); 134 hval = NEXT (hval, ip);
97 hslot = htab + IDX (hval); 135 hslot = htab + IDX (hval);
98 ref = *hslot; *hslot = ip; 136 ref = *hslot; *hslot = ip;
99 137
100 if (1 138 if (1
101#if INIT_HTAB && !USE_MEMCPY 139#if INIT_HTAB && !USE_MEMCPY
102 && ref < ip /* the next test will actually take care of this, but it is faster */ 140 && ref < ip /* the next test will actually take care of this, but this is faster */
103#endif 141#endif
104 && (off = ip - ref - 1) < MAX_OFF 142 && (off = ip - ref - 1) < MAX_OFF
105 && ip + 4 < in_end 143 && ip + 4 < in_end
106 && ref > (u8 *)in_data 144 && ref > (u8 *)in_data
107#if STRICT_ALIGN 145#if STRICT_ALIGN
108 && ref[0] == ip[0] 146 && ref[0] == ip[0]
109 && ref[1] == ip[1] 147 && ref[1] == ip[1]
110 && ref[2] == ip[2] 148 && ref[2] == ip[2]
111#else 149#else
112 && *(u16 *)ref == *(u16 *)ip 150 && *(u16 *)ref == *(u16 *)ip
113 && ref[2] == ip[2] 151 && ref[2] == ip[2]
114#endif 152#endif
115 ) 153 )
116 { 154 {
117 /* match found at *ref++ */ 155 /* match found at *ref++ */
118 unsigned int len = 2; 156 unsigned int len = 2;
119 unsigned int maxlen = in_end - ip - len; 157 unsigned int maxlen = in_end - ip - len;
120 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; 158 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen;
121 159
122 do 160 if (expect_false (op + lit + 1 + 3 >= out_end))
123 len++; 161 return 0;
162
163 if (lit)
164 {
165 *op++ = lit - 1;
166 lit = -lit;
167 do
168 *op++ = ip[lit];
169 while (++lit);
170 }
171
172 for (;;)
173 {
174 if (expect_true (ip < in_end - 2 - 8 && maxlen > 8))
175 {
176 len++; if (ref [len] != ip [len]) break;
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 }
185
186 do
187 len++;
124 while (len < maxlen && ref[len] == ip[len]); 188 while (len < maxlen && ref[len] == ip[len]);
125 189
126 if (op + lit + 1 + 3 >= out_end) 190 break;
127 return 0; 191 }
128 192
129 if (lit) 193 len -= 2;
130 { 194 ip++;
131 *op++ = lit - 1;
132 lit = -lit;
133 do
134 *op++ = ip[lit];
135 while (++lit);
136 }
137 195
138 len -= 2; 196 if (len < 7)
139 ip++; 197 {
140
141 if (len < 7)
142 {
143 *op++ = (off >> 8) + (len << 5); 198 *op++ = (off >> 8) + (len << 5);
144 } 199 }
145 else 200 else
146 { 201 {
147 *op++ = (off >> 8) + ( 7 << 5); 202 *op++ = (off >> 8) + ( 7 << 5);
148 *op++ = len - 7; 203 *op++ = len - 7;
149 } 204 }
150 205
151 *op++ = off; 206 *op++ = off;
152 207
153#if ULTRA_FAST 208#if ULTRA_FAST || VERY_FAST
154 ip += len; 209 ip += len;
210#if VERY_FAST && !ULTRA_FAST
211 --ip;
212#endif
155 hval = FRST (ip); 213 hval = FRST (ip);
214
156 hval = NEXT (hval, ip); 215 hval = NEXT (hval, ip);
157 htab[IDX (hval)] = ip; 216 htab[IDX (hval)] = ip;
158 ip++; 217 ip++;
159#else 218
160 do 219#if VERY_FAST && !ULTRA_FAST
161 {
162 hval = NEXT (hval, ip); 220 hval = NEXT (hval, ip);
163 htab[IDX (hval)] = ip; 221 htab[IDX (hval)] = ip;
222 ip++;
223#endif
224#else
225 do
226 {
227 hval = NEXT (hval, ip);
228 htab[IDX (hval)] = ip;
229 ip++;
230 }
231 while (len--);
232#endif
233 continue;
234 }
235 }
236 else if (expect_false (ip == in_end))
237 break;
238
239 /* one more literal byte we must copy */
240 lit++;
164 ip++; 241 ip++;
165 }
166 while (len--);
167#endif
168 }
169 else
170 {
171 /* one more literal byte we must copy */
172 lit++;
173 ip++;
174 242
175 if (lit == MAX_LIT) 243 if (expect_false (lit == MAX_LIT))
176 { 244 {
177 if (op + 1 + MAX_LIT >= out_end) 245 if (op + 1 + MAX_LIT >= out_end)
178 return 0; 246 return 0;
179 247
180 *op++ = MAX_LIT - 1; 248 *op++ = MAX_LIT - 1;
181#if USE_MEMCPY 249
182 memcpy (op, ip - MAX_LIT, MAX_LIT); 250#ifdef lzf_movsb
183 op += MAX_LIT; 251 ip -= lit;
184 lit = 0; 252 lzf_movsb (op, ip, lit);
185#else 253#else
186 lit = -lit; 254 lit = -lit;
187 do 255 do
188 *op++ = ip[lit]; 256 *op++ = ip[lit];
189 while (++lit); 257 while (++lit);
190#endif 258#endif
191 } 259 }
192 }
193 } 260 }
194 while (ip < in_end);
195 261
196 if (lit) 262 if (lit)
197 { 263 {
198 if (op + lit + 1 >= out_end) 264 if (op + lit + 1 >= out_end)
199 return 0; 265 return 0;
200 266
201 *op++ = lit - 1; 267 *op++ = lit - 1;
268#ifdef lzf_movsb
269 ip -= lit;
270 lzf_movsb (op, ip, lit);
271#else
202 lit = -lit; 272 lit = -lit;
203 do 273 do
204 *op++ = ip[lit]; 274 *op++ = ip[lit];
205 while (++lit); 275 while (++lit);
276#endif
206 } 277 }
207 278
208 return op - (u8 *) out_data; 279 return op - (u8 *) out_data;
209} 280}
281

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines