1 | /* |
1 | /* |
2 | * Copyright (c) 2000-2002 Marc Alexander Lehmann <pcg@goof.com> |
2 | * Copyright (c) 2000-2010,2012 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 | * |
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 | * |
|
|
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- |
21 | * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
18 | * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, |
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 | #if __GNUC__ >= 3 |
57 | #define MAX_OFF (1 << 13) |
75 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
58 | #define MAX_REF ((1 << 8) + (1 << 3)) |
76 | # define inline inline |
|
|
77 | #else |
|
|
78 | # define expect(expr,value) (expr) |
|
|
79 | # define inline static |
|
|
80 | #endif |
|
|
81 | |
|
|
82 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
83 | #define expect_true(expr) expect ((expr) != 0, 1) |
59 | |
84 | |
60 | /* |
85 | /* |
61 | * compressed format |
86 | * compressed format |
62 | * |
87 | * |
63 | * 000LLLLL <L+1> ; literal |
88 | * 000LLLLL <L+1> ; literal, L+1=1..33 octets |
64 | * LLLOOOOO oooooooo ; backref L |
89 | * LLLooooo oooooooo ; backref L+1=1..7 octets, o+1=1..4096 offset |
65 | * 111OOOOO LLLLLLLL oooooooo ; backref L+7 |
90 | * 111ooooo LLLLLLLL oooooooo ; backref L+8 octets, o+1=1..4096 offset |
66 | * |
91 | * |
67 | */ |
92 | */ |
68 | |
93 | |
69 | unsigned int |
94 | unsigned int |
70 | lzf_compress (const void *const in_data, unsigned int in_len, |
95 | lzf_compress (const void *const in_data, unsigned int in_len, |
71 | void *out_data, unsigned int out_len) |
96 | void *out_data, unsigned int out_len |
|
|
97 | #if LZF_STATE_ARG |
|
|
98 | , LZF_STATE htab |
|
|
99 | #endif |
|
|
100 | ) |
72 | { |
101 | { |
73 | const u8 *htab[HSIZE]; |
102 | #if !LZF_STATE_ARG |
74 | const u8 **hslot; |
103 | LZF_STATE htab; |
|
|
104 | #endif |
75 | const u8 *ip = in_data; |
105 | const u8 *ip = (const u8 *)in_data; |
76 | u8 *op = out_data; |
106 | u8 *op = (u8 *)out_data; |
77 | const u8 *in_end = ip + in_len; |
107 | const u8 *in_end = ip + in_len; |
78 | u8 *out_end = op + out_len; |
108 | u8 *out_end = op + out_len; |
79 | const u8 *ref; |
109 | const u8 *ref; |
80 | |
110 | |
81 | unsigned int hval = FRST (ip); |
111 | /* off requires a type wide enough to hold a general pointer difference. |
|
|
112 | * ISO C doesn't have that (size_t might not be enough and ptrdiff_t only |
|
|
113 | * works for differences within a single object). We also assume that |
|
|
114 | * no bit pattern traps. Since the only platform that is both non-POSIX |
|
|
115 | * and fails to support both assumptions is windows 64 bit, we make a |
|
|
116 | * special workaround for it. |
|
|
117 | */ |
|
|
118 | #if defined (_WIN32) && defined (_M_X64) |
|
|
119 | /* workaround for missing POSIX compliance */ |
|
|
120 | #if __GNUC__ |
|
|
121 | unsigned long long off; |
|
|
122 | #else |
|
|
123 | unsigned __int64 off; |
|
|
124 | #endif |
|
|
125 | #else |
82 | unsigned long off; |
126 | unsigned long off; |
83 | int lit = 0; |
127 | #endif |
|
|
128 | unsigned int hval; |
|
|
129 | int lit; |
|
|
130 | |
|
|
131 | if (!in_len || !out_len) |
|
|
132 | return 0; |
84 | |
133 | |
85 | #if INIT_HTAB |
134 | #if INIT_HTAB |
86 | # if USE_MEMCPY |
|
|
87 | memset (htab, 0, sizeof (htab)); |
135 | memset (htab, 0, sizeof (LZF_STATE)); |
88 | # else |
|
|
89 | for (hslot = htab; hslot < htab + HSIZE; hslot++) |
|
|
90 | *hslot++ = ip; |
|
|
91 | # endif |
136 | #endif |
92 | #endif |
|
|
93 | |
137 | |
94 | do |
138 | lit = 0; op++; /* start run */ |
|
|
139 | |
|
|
140 | hval = FRST (ip); |
|
|
141 | while (ip < in_end - 2) |
95 | { |
142 | { |
|
|
143 | LZF_HSLOT *hslot; |
|
|
144 | |
96 | hval = NEXT (hval, ip); |
145 | hval = NEXT (hval, ip); |
97 | hslot = htab + IDX (hval); |
146 | hslot = htab + IDX (hval); |
98 | ref = *hslot; *hslot = ip; |
147 | ref = *hslot + LZF_HSLOT_BIAS; *hslot = ip - LZF_HSLOT_BIAS; |
99 | |
148 | |
100 | if (1 |
149 | if (1 |
101 | #if INIT_HTAB && !USE_MEMCPY |
150 | #if INIT_HTAB |
102 | && ref < ip /* the next test will actually take care of this, but it is faster */ |
151 | && ref < ip /* the next test will actually take care of this, but this is faster */ |
103 | #endif |
152 | #endif |
104 | && (off = ip - ref - 1) < MAX_OFF |
153 | && (off = ip - ref - 1) < LZF_MAX_OFF |
105 | && ip + 4 < in_end |
|
|
106 | && ref > (u8 *)in_data |
154 | && ref > (u8 *)in_data |
|
|
155 | && ref[2] == ip[2] |
107 | #if STRICT_ALIGN |
156 | #if STRICT_ALIGN |
108 | && ref[0] == ip[0] |
157 | && ((ref[1] << 8) | ref[0]) == ((ip[1] << 8) | ip[0]) |
109 | && ref[1] == ip[1] |
|
|
110 | && ref[2] == ip[2] |
|
|
111 | #else |
158 | #else |
112 | && *(u16 *)ref == *(u16 *)ip |
159 | && *(u16 *)ref == *(u16 *)ip |
113 | && ref[2] == ip[2] |
|
|
114 | #endif |
160 | #endif |
115 | ) |
161 | ) |
116 | { |
162 | { |
117 | /* match found at *ref++ */ |
163 | /* match found at *ref++ */ |
118 | unsigned int len = 2; |
164 | unsigned int len = 2; |
119 | unsigned int maxlen = in_end - ip - len; |
165 | unsigned int maxlen = in_end - ip - len; |
120 | maxlen = maxlen > MAX_REF ? MAX_REF : maxlen; |
166 | maxlen = maxlen > LZF_MAX_REF ? LZF_MAX_REF : maxlen; |
121 | |
167 | |
122 | do |
168 | if (expect_false (op + 3 + 1 >= out_end)) /* first a faster conservative test */ |
123 | len++; |
169 | if (op - !lit + 3 + 1 >= out_end) /* second the exact but rare test */ |
|
|
170 | return 0; |
|
|
171 | |
|
|
172 | op [- lit - 1] = lit - 1; /* stop run */ |
|
|
173 | op -= !lit; /* undo run if length is zero */ |
|
|
174 | |
|
|
175 | for (;;) |
|
|
176 | { |
|
|
177 | if (expect_true (maxlen > 16)) |
|
|
178 | { |
|
|
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 | |
|
|
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 | len++; if (ref [len] != ip [len]) break; |
|
|
188 | |
|
|
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 | len++; if (ref [len] != ip [len]) break; |
|
|
193 | |
|
|
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 | len++; if (ref [len] != ip [len]) break; |
|
|
198 | } |
|
|
199 | |
|
|
200 | do |
|
|
201 | len++; |
124 | while (len < maxlen && ref[len] == ip[len]); |
202 | while (len < maxlen && ref[len] == ip[len]); |
125 | |
203 | |
126 | if (op + lit + 1 + 3 >= out_end) |
204 | break; |
127 | return 0; |
205 | } |
128 | |
206 | |
129 | if (lit) |
207 | len -= 2; /* len is now #octets - 1 */ |
130 | { |
208 | ip++; |
131 | *op++ = lit - 1; |
|
|
132 | lit = -lit; |
|
|
133 | do |
|
|
134 | *op++ = ip[lit]; |
|
|
135 | while (++lit); |
|
|
136 | } |
|
|
137 | |
209 | |
138 | len -= 2; |
|
|
139 | ip++; |
|
|
140 | |
|
|
141 | if (len < 7) |
210 | if (len < 7) |
142 | { |
211 | { |
143 | *op++ = (off >> 8) + (len << 5); |
212 | *op++ = (off >> 8) + (len << 5); |
144 | } |
213 | } |
145 | else |
214 | else |
146 | { |
215 | { |
147 | *op++ = (off >> 8) + ( 7 << 5); |
216 | *op++ = (off >> 8) + ( 7 << 5); |
148 | *op++ = len - 7; |
217 | *op++ = len - 7; |
149 | } |
218 | } |
150 | |
219 | |
151 | *op++ = off; |
220 | *op++ = off; |
152 | |
221 | |
153 | #if ULTRA_FAST |
222 | lit = 0; op++; /* start run */ |
|
|
223 | |
154 | ip += len; |
224 | ip += len + 1; |
|
|
225 | |
|
|
226 | if (expect_false (ip >= in_end - 2)) |
|
|
227 | break; |
|
|
228 | |
|
|
229 | #if ULTRA_FAST || VERY_FAST |
|
|
230 | --ip; |
|
|
231 | # if VERY_FAST && !ULTRA_FAST |
|
|
232 | --ip; |
|
|
233 | # endif |
155 | hval = FRST (ip); |
234 | hval = FRST (ip); |
|
|
235 | |
156 | hval = NEXT (hval, ip); |
236 | hval = NEXT (hval, ip); |
157 | htab[IDX (hval)] = ip; |
237 | htab[IDX (hval)] = ip - LZF_HSLOT_BIAS; |
158 | ip++; |
238 | ip++; |
159 | #else |
239 | |
160 | do |
240 | # if VERY_FAST && !ULTRA_FAST |
161 | { |
|
|
162 | hval = NEXT (hval, ip); |
241 | hval = NEXT (hval, ip); |
163 | htab[IDX (hval)] = ip; |
242 | htab[IDX (hval)] = ip - LZF_HSLOT_BIAS; |
164 | ip++; |
243 | ip++; |
165 | } |
|
|
166 | while (len--); |
|
|
167 | #endif |
244 | # endif |
168 | } |
245 | #else |
|
|
246 | ip -= len + 1; |
|
|
247 | |
|
|
248 | do |
|
|
249 | { |
|
|
250 | hval = NEXT (hval, ip); |
|
|
251 | htab[IDX (hval)] = ip - LZF_HSLOT_BIAS; |
|
|
252 | ip++; |
|
|
253 | } |
|
|
254 | while (len--); |
|
|
255 | #endif |
|
|
256 | } |
169 | else |
257 | else |
170 | { |
258 | { |
171 | /* one more literal byte we must copy */ |
259 | /* one more literal byte we must copy */ |
172 | lit++; |
260 | if (expect_false (op >= out_end)) |
173 | ip++; |
261 | return 0; |
174 | |
262 | |
175 | if (lit == MAX_LIT) |
263 | lit++; *op++ = *ip++; |
176 | { |
|
|
177 | if (op + 1 + MAX_LIT >= out_end) |
|
|
178 | return 0; |
|
|
179 | |
264 | |
180 | *op++ = MAX_LIT - 1; |
265 | if (expect_false (lit == LZF_MAX_LIT)) |
181 | #if USE_MEMCPY |
266 | { |
182 | memcpy (op, ip - MAX_LIT, MAX_LIT); |
267 | op [- lit - 1] = lit - 1; /* stop run */ |
183 | op += MAX_LIT; |
268 | lit = 0; op++; /* start run */ |
184 | lit = 0; |
269 | } |
185 | #else |
270 | } |
186 | lit = -lit; |
|
|
187 | do |
|
|
188 | *op++ = ip[lit]; |
|
|
189 | while (++lit); |
|
|
190 | #endif |
|
|
191 | } |
|
|
192 | } |
|
|
193 | } |
271 | } |
|
|
272 | |
|
|
273 | if (op + 3 > out_end) /* at most 3 bytes can be missing here */ |
|
|
274 | return 0; |
|
|
275 | |
194 | while (ip < in_end); |
276 | while (ip < in_end) |
195 | |
|
|
196 | if (lit) |
|
|
197 | { |
277 | { |
198 | if (op + lit + 1 >= out_end) |
278 | lit++; *op++ = *ip++; |
199 | return 0; |
|
|
200 | |
279 | |
201 | *op++ = lit - 1; |
280 | if (expect_false (lit == LZF_MAX_LIT)) |
202 | lit = -lit; |
281 | { |
203 | do |
282 | op [- lit - 1] = lit - 1; /* stop run */ |
204 | *op++ = ip[lit]; |
283 | lit = 0; op++; /* start run */ |
205 | while (++lit); |
284 | } |
206 | } |
285 | } |
207 | |
286 | |
|
|
287 | op [- lit - 1] = lit - 1; /* end run */ |
|
|
288 | op -= !lit; /* undo run if length is zero */ |
|
|
289 | |
208 | return op - (u8 *) out_data; |
290 | return op - (u8 *)out_data; |
209 | } |
291 | } |
|
|
292 | |