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 1.2 by root, Thu Sep 5 03:31:00 2002 UTC vs.
Revision 1.14 by root, Thu Jun 21 22:11:34 2007 UTC

1/* 1/*
2 * Copyright (c) 2000 Marc Alexander Lehmann <pcg@goof.com> 2 * Copyright (c) 2000-2005 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,
22 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 22 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
23 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 23 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
24 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH- 24 * 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 25 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
26 * OF THE POSSIBILITY OF SUCH DAMAGE. 26 * OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * 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
30 * provisions of the GPL are applicable instead of the above. If you wish to
31 * allow the use of your version of this file only under the terms of the
32 * GPL and not to allow others to use your version of this file under the
33 * BSD license, indicate your decision by deleting the provisions above and
34 * replace them with the notice and other provisions required by the GPL. If
35 * you do not delete the provisions above, a recipient may use your version
36 * of this file under either the BSD or the GPL.
27 */ 37 */
28 38
29#include "lzfP.h" 39#include "lzfP.h"
30 40
31#define HSIZE (1 << (HLOG)) 41#define HSIZE (1 << (HLOG))
34 * don't play with this unless you benchmark! 44 * don't play with this unless you benchmark!
35 * decompression is not dependent on the hash function 45 * decompression is not dependent on the hash function
36 * the hashing function might seem strange, just believe me 46 * the hashing function might seem strange, just believe me
37 * it works ;) 47 * it works ;)
38 */ 48 */
49#ifndef FRST
39#define FRST(p) (((p[0]) << 8) + p[1]) 50# define FRST(p) (((p[0]) << 8) | p[1])
40#define NEXT(v,p) (((v) << 8) + p[2]) 51# define NEXT(v,p) (((v) << 8) | p[2])
41#define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) + h*3) & (HSIZE - 1)) 52# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
53/*# define IDX(h) ((ip[0] * 121 ^ ip[1] * 33 ^ ip[2] * 1) & (HSIZE-1))*/
54#endif
42/* 55/*
43 * IDX works because it is very similar to a multiplicative hash, e.g. 56 * IDX works because it is very similar to a multiplicative hash, e.g.
44 * (h * 57321 >> (3*8 - HLOG)) 57 * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1))
58 * the latter is also quite fast on newer CPUs, and compresses similarly.
59 *
45 * the next one is also quite good, albeit slow ;) 60 * the next one is also quite good, albeit slow ;)
46 * (int)(cos(h & 0xffffff) * 1e6) 61 * (int)(cos(h & 0xffffff) * 1e6)
47 */ 62 */
48 63
49#if 0 64#if 0
50/* original lzv-like hash function */ 65/* original lzv-like hash function, much worse and thus slower */
51# define FRST(p) (p[0] << 5) ^ p[1] 66# define FRST(p) (p[0] << 5) ^ p[1]
52# define NEXT(v,p) ((v) << 5) ^ p[2] 67# define NEXT(v,p) ((v) << 5) ^ p[2]
53# define IDX(h) ((h) & (HSIZE - 1)) 68# define IDX(h) ((h) & (HSIZE - 1))
54#endif 69#endif
55 70
56#define MAX_LIT (1 << 5) 71#define MAX_LIT (1 << 5)
57#define MAX_OFF (1 << 13) 72#define MAX_OFF (1 << 13)
58#define MAX_REF ((1 << 8) + (1 << 3)) 73#define MAX_REF ((1 << 8) + (1 << 3))
74
75#if (__i386 || __amd64) && __GNUC__ >= 3
76# define lzf_movsb(dst, src, len) \
77 asm ("rep movsb" \
78 : "=D" (dst), "=S" (src), "=c" (len) \
79 : "0" (dst), "1" (src), "2" (len));
80#endif
59 81
60/* 82/*
61 * compressed format 83 * compressed format
62 * 84 *
63 * 000LLLLL <L+1> ; literal 85 * 000LLLLL <L+1> ; literal
64 * LLLOOOOO oooooooo ; backref L 86 * LLLooooo oooooooo ; backref L
65 * 111OOOOO LLLLLLLL oooooooo ; backref L+7 87 * 111ooooo LLLLLLLL oooooooo ; backref L+7
66 * 88 *
67 */ 89 */
68 90
69unsigned int 91unsigned int
70lzf_compress (const void *const in_data, unsigned int in_len, 92lzf_compress (const void *const in_data, unsigned int in_len,
71 void *out_data, unsigned int out_len) 93 void *out_data, unsigned int out_len
94#if LZF_STATE_ARG
95 , LZF_STATE htab
96#endif
97 )
72{ 98{
73 const u8 *htab[HSIZE]; 99#if !LZF_STATE_ARG
100 LZF_STATE htab;
101#endif
74 const u8 **hslot; 102 const u8 **hslot;
75 const u8 *ip = in_data; 103 const u8 *ip = (const u8 *)in_data;
76 u8 *op = out_data; 104 u8 *op = (u8 *)out_data;
77 const u8 *in_end = ip + in_len; 105 const u8 *in_end = ip + in_len;
78 u8 *out_end = op + out_len; 106 u8 *out_end = op + out_len;
79 const u8 *ref; 107 const u8 *ref;
80 108
81 unsigned int hval = FRST (ip); 109 unsigned int hval = FRST (ip);
82 unsigned long off; 110 unsigned long off;
83 int lit = 0; 111 int lit = 0;
84 112
85#if INIT_HTAB 113#if INIT_HTAB
86# if USE_MEMCPY
87 memset (htab, 0, sizeof (htab)); 114 memset (htab, 0, sizeof (htab));
88# else 115# if 0
89 for (hslot = htab; hslot < htab + HSIZE; hslot++) 116 for (hslot = htab; hslot < htab + HSIZE; hslot++)
90 *hslot++ = ip; 117 *hslot++ = ip;
91# endif 118# endif
92#endif 119#endif
93 120
94 do 121 for (;;)
95 { 122 {
123 if (ip < in_end - 2)
124 {
96 hval = NEXT (hval, ip); 125 hval = NEXT (hval, ip);
97 hslot = htab + IDX (hval); 126 hslot = htab + IDX (hval);
98 ref = *hslot; *hslot = ip; 127 ref = *hslot; *hslot = ip;
99 128
100 if (1 129 if (1
101#if INIT_HTAB && !USE_MEMCPY 130#if INIT_HTAB && !USE_MEMCPY
102 && ref < ip /* the next test will actually take care of this, but it is faster */ 131 && ref < ip /* the next test will actually take care of this, but this is faster */
103#endif 132#endif
104 && (off = ip - ref - 1) < MAX_OFF 133 && (off = ip - ref - 1) < MAX_OFF
105 && ip + 4 < in_end 134 && ip + 4 < in_end
106 && ref > (u8 *)in_data 135 && ref > (u8 *)in_data
107#if STRICT_ALIGN 136#if STRICT_ALIGN
108 && ref[0] == ip[0] 137 && ref[0] == ip[0]
109 && ref[1] == ip[1] 138 && ref[1] == ip[1]
110 && ref[2] == ip[2] 139 && ref[2] == ip[2]
111#else 140#else
112 && *(u16 *)ref == *(u16 *)ip 141 && *(u16 *)ref == *(u16 *)ip
113 && ref[2] == ip[2] 142 && ref[2] == ip[2]
114#endif 143#endif
115 ) 144 )
116 { 145 {
117 /* match found at *ref++ */ 146 /* match found at *ref++ */
118 unsigned int len = 2; 147 unsigned int len = 2;
119 unsigned int maxlen = in_end - ip - len; 148 unsigned int maxlen = in_end - ip - len;
120 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; 149 maxlen = maxlen > MAX_REF ? MAX_REF : maxlen;
121 150
122 do 151 if (op + lit + 1 + 3 >= out_end)
123 len++; 152 return 0;
153
154 do
155 len++;
124 while (len < maxlen && ref[len] == ip[len]); 156 while (len < maxlen && ref[len] == ip[len]);
125 157
126 if (op + lit + 1 + 3 >= out_end) 158 if (lit)
127 return 0; 159 {
128
129 if (lit)
130 {
131 *op++ = lit - 1; 160 *op++ = lit - 1;
132 lit = -lit; 161 lit = -lit;
133 do 162 do
134 *op++ = ip[lit]; 163 *op++ = ip[lit];
135 while (++lit); 164 while (++lit);
136 } 165 }
137 166
138 len -= 2; 167 len -= 2;
139 ip++; 168 ip++;
140 169
141 if (len < 7) 170 if (len < 7)
142 { 171 {
143 *op++ = (off >> 8) + (len << 5); 172 *op++ = (off >> 8) + (len << 5);
144 } 173 }
145 else 174 else
146 { 175 {
147 *op++ = (off >> 8) + ( 7 << 5); 176 *op++ = (off >> 8) + ( 7 << 5);
148 *op++ = len - 7; 177 *op++ = len - 7;
149 } 178 }
150 179
151 *op++ = off; 180 *op++ = off;
152 181
153#if ULTRA_FAST 182#if ULTRA_FAST || VERY_FAST
154 ip += len; 183 ip += len;
184#if VERY_FAST && !ULTRA_FAST
185 --ip;
186#endif
155 hval = FRST (ip); 187 hval = FRST (ip);
188
156 hval = NEXT (hval, ip); 189 hval = NEXT (hval, ip);
157 htab[IDX (hval)] = ip; 190 htab[IDX (hval)] = ip;
158 ip++; 191 ip++;
159#else 192
160 do 193#if VERY_FAST && !ULTRA_FAST
161 {
162 hval = NEXT (hval, ip); 194 hval = NEXT (hval, ip);
163 htab[IDX (hval)] = ip; 195 htab[IDX (hval)] = ip;
196 ip++;
197#endif
198#else
199 do
200 {
201 hval = NEXT (hval, ip);
202 htab[IDX (hval)] = ip;
203 ip++;
204 }
205 while (len--);
206#endif
207 continue;
208 }
209 }
210 else if (ip == in_end)
211 break;
212
213 /* one more literal byte we must copy */
214 lit++;
164 ip++; 215 ip++;
165 }
166 while (len--);
167#endif
168 }
169 else
170 {
171 /* one more literal byte we must copy */
172 lit++;
173 ip++;
174 216
175 if (lit == MAX_LIT) 217 if (lit == MAX_LIT)
176 { 218 {
177 if (op + 1 + MAX_LIT >= out_end) 219 if (op + 1 + MAX_LIT >= out_end)
178 return 0; 220 return 0;
179 221
180 *op++ = MAX_LIT - 1; 222 *op++ = MAX_LIT - 1;
181#if USE_MEMCPY 223
182 memcpy (op, ip - MAX_LIT, MAX_LIT); 224#ifdef lzf_movsb
183 op += MAX_LIT; 225 ip -= lit;
184 lit = 0; 226 lzf_movsb (op, ip, lit);
185#else 227#else
186 lit = -lit; 228 lit = -lit;
187 do 229 do
188 *op++ = ip[lit]; 230 *op++ = ip[lit];
189 while (++lit); 231 while (++lit);
190#endif 232#endif
191 } 233 }
192 }
193 } 234 }
194 while (ip < in_end);
195 235
196 if (lit) 236 if (lit)
197 { 237 {
198 if (op + lit + 1 >= out_end) 238 if (op + lit + 1 >= out_end)
199 return 0; 239 return 0;
200 240
201 *op++ = lit - 1; 241 *op++ = lit - 1;
242#ifdef lzf_movsb
243 ip -= lit;
244 lzf_movsb (op, ip, lit);
245#else
202 lit = -lit; 246 lit = -lit;
203 do 247 do
204 *op++ = ip[lit]; 248 *op++ = ip[lit];
205 while (++lit); 249 while (++lit);
250#endif
206 } 251 }
207 252
208 return op - (u8 *) out_data; 253 return op - (u8 *) out_data;
209} 254}
255

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines