Line data Source code
1 : /*
2 : * Copyright (c) 2000-2010 Marc Alexander Lehmann <schmorp@schmorp.de>
3 : *
4 : * Redistribution and use in source and binary forms, with or without modifica-
5 : * tion, are permitted provided that the following conditions are met:
6 : *
7 : * 1. Redistributions of source code must retain the above copyright notice,
8 : * this list of conditions and the following disclaimer.
9 : *
10 : * 2. Redistributions in binary form must reproduce the above copyright
11 : * notice, this list of conditions and the following disclaimer in the
12 : * documentation and/or other materials provided with the distribution.
13 : *
14 : * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
15 : * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
16 : * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO
17 : * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
18 : * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
19 : * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
20 : * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
21 : * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
22 : * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
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.
35 : */
36 :
37 : #include "lzfP.h"
38 :
39 : #define HSIZE (1 << (HLOG))
40 :
41 : /*
42 : * don't play with this unless you benchmark!
43 : * the data format is not dependent on the hash function.
44 : * the hash function might seem strange, just believe me,
45 : * it works ;)
46 : */
47 : #ifndef FRST
48 : # define FRST(p) (((p[0]) << 8) | p[1])
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
55 : # define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1))
56 : # endif
57 : #endif
58 : /*
59 : * IDX works because it is very similar to a multiplicative hash, e.g.
60 : * ((h * 57321 >> (3*8 - HLOG)) & (HSIZE - 1))
61 : * the latter is also quite fast on newer CPUs, and compresses similarly.
62 : *
63 : * the next one is also quite good, albeit slow ;)
64 : * (int)(cos(h & 0xffffff) * 1e6)
65 : */
66 :
67 : #if 0
68 : /* original lzv-like hash function, much worse and thus slower */
69 : # define FRST(p) (p[0] << 5) ^ p[1]
70 : # define NEXT(v,p) ((v) << 5) ^ p[2]
71 : # define IDX(h) ((h) & (HSIZE - 1))
72 : #endif
73 :
74 : #define MAX_LIT (1 << 5)
75 : #define MAX_OFF (1 << 13)
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)
88 :
89 : /*
90 : * compressed format
91 : *
92 : * 000LLLLL <L+1> ; literal, L+1=1..33 octets
93 : * LLLooooo oooooooo ; backref L+1=1..7 octets, o+1=1..4096 offset
94 : * 111ooooo LLLLLLLL oooooooo ; backref L+8 octets, o+1=1..4096 offset
95 : *
96 : */
97 :
98 : unsigned int
99 74 : lzf_compress (const void *const in_data, unsigned int in_len,
100 : void *out_data, unsigned int out_len
101 : #if LZF_STATE_ARG
102 : , LZF_STATE htab
103 : #endif
104 : )
105 : {
106 : #if !LZF_STATE_ARG
107 : LZF_STATE htab;
108 : #endif
109 74 : const u8 *ip = (const u8 *)in_data;
110 74 : u8 *op = (u8 *)out_data;
111 74 : const u8 *in_end = ip + in_len;
112 74 : u8 *out_end = op + out_len;
113 : const u8 *ref;
114 :
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 long long off; /* workaround for missing POSIX compliance */
124 : #else
125 : unsigned long off;
126 : #endif
127 : unsigned int hval;
128 : int lit;
129 :
130 74 : if (!in_len || !out_len)
131 0 : return 0;
132 :
133 : #if INIT_HTAB
134 : memset (htab, 0, sizeof (htab));
135 : #endif
136 :
137 74 : lit = 0; op++; /* start run */
138 :
139 74 : hval = FRST (ip);
140 293195052 : while (ip < in_end - 2)
141 : {
142 : LZF_HSLOT *hslot;
143 :
144 293194960 : hval = NEXT (hval, ip);
145 293194960 : hslot = htab + IDX (hval);
146 293194960 : ref = *hslot + LZF_HSLOT_BIAS; *hslot = ip - LZF_HSLOT_BIAS;
147 :
148 293194960 : 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 293194960 : && (off = ip - ref - 1) < MAX_OFF
153 61201757 : && ref > (u8 *)in_data
154 61200832 : && ref[2] == ip[2]
155 : #if STRICT_ALIGN
156 : && ((ref[1] << 8) | ref[0]) == ((ip[1] << 8) | ip[0])
157 : #else
158 33328530 : && *(u16 *)ref == *(u16 *)ip
159 : #endif
160 : )
161 33327944 : {
162 : /* match found at *ref++ */
163 33328000 : unsigned int len = 2;
164 33328000 : unsigned int maxlen = in_end - ip - len;
165 33328000 : maxlen = maxlen > MAX_REF ? MAX_REF : maxlen;
166 :
167 33328000 : if (expect_false (op + 3 + 1 >= out_end)) /* first a faster conservative test */
168 0 : if (op - !lit + 3 + 1 >= out_end) /* second the exact but rare test */
169 0 : return 0;
170 :
171 33328000 : op [- lit - 1] = lit - 1; /* stop run */
172 33328000 : op -= !lit; /* undo run if length is zero */
173 :
174 : for (;;)
175 : {
176 33328000 : if (expect_true (maxlen > 16))
177 : {
178 33327886 : len++; if (ref [len] != ip [len]) break;
179 16469008 : len++; if (ref [len] != ip [len]) break;
180 6775012 : len++; if (ref [len] != ip [len]) break;
181 5768808 : len++; if (ref [len] != ip [len]) break;
182 :
183 4958614 : len++; if (ref [len] != ip [len]) break;
184 3647494 : len++; if (ref [len] != ip [len]) break;
185 2679034 : len++; if (ref [len] != ip [len]) break;
186 2466468 : len++; if (ref [len] != ip [len]) break;
187 :
188 2301162 : len++; if (ref [len] != ip [len]) break;
189 1987524 : len++; if (ref [len] != ip [len]) break;
190 1757540 : len++; if (ref [len] != ip [len]) break;
191 1681662 : len++; if (ref [len] != ip [len]) break;
192 :
193 1597304 : len++; if (ref [len] != ip [len]) break;
194 956852 : len++; if (ref [len] != ip [len]) break;
195 805590 : len++; if (ref [len] != ip [len]) break;
196 746104 : len++; if (ref [len] != ip [len]) break;
197 : }
198 :
199 : do
200 17384454 : len++;
201 17384454 : while (len < maxlen && ref[len] == ip[len]);
202 :
203 689460 : break;
204 : }
205 :
206 33328000 : len -= 2; /* len is now #octets - 1 */
207 33328000 : ip++;
208 :
209 33328000 : if (len < 7)
210 : {
211 30648938 : *op++ = (off >> 8) + (len << 5);
212 : }
213 : else
214 : {
215 2679062 : *op++ = (off >> 8) + ( 7 << 5);
216 2679062 : *op++ = len - 7;
217 : }
218 :
219 33328000 : *op++ = off;
220 :
221 33328000 : lit = 0; op++; /* start run */
222 :
223 33328000 : ip += len + 1;
224 :
225 33328000 : if (expect_false (ip >= in_end - 2))
226 56 : break;
227 :
228 : #if ULTRA_FAST || VERY_FAST
229 33327944 : --ip;
230 : # if VERY_FAST && !ULTRA_FAST
231 33327944 : --ip;
232 : # endif
233 33327944 : hval = FRST (ip);
234 :
235 33327944 : hval = NEXT (hval, ip);
236 33327944 : htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
237 33327944 : ip++;
238 :
239 : # if VERY_FAST && !ULTRA_FAST
240 33327944 : hval = NEXT (hval, ip);
241 33327944 : htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
242 33327944 : 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;
251 : ip++;
252 : }
253 : while (len--);
254 : #endif
255 : }
256 : else
257 : {
258 : /* one more literal byte we must copy */
259 259866960 : if (expect_false (op >= out_end))
260 0 : return 0;
261 :
262 259866960 : lit++; *op++ = *ip++;
263 :
264 259866960 : if (expect_false (lit == MAX_LIT))
265 : {
266 5785126 : op [- lit - 1] = lit - 1; /* stop run */
267 5785126 : lit = 0; op++; /* start run */
268 : }
269 : }
270 : }
271 :
272 74 : if (op + 3 > out_end) /* at most 3 bytes can be missing here */
273 0 : return 0;
274 :
275 288 : while (ip < in_end)
276 : {
277 140 : lit++; *op++ = *ip++;
278 :
279 140 : if (expect_false (lit == MAX_LIT))
280 : {
281 0 : op [- lit - 1] = lit - 1; /* stop run */
282 0 : lit = 0; op++; /* start run */
283 : }
284 : }
285 :
286 74 : op [- lit - 1] = lit - 1; /* end run */
287 74 : op -= !lit; /* undo run if length is zero */
288 :
289 74 : return op - (u8 *)out_data;
290 : }
291 :
|
