| 1 |
/* |
| 2 |
* Copyright (c) 2000-2010,2012 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 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 |
| 56 |
# define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) - h*5) & (HSIZE - 1)) |
| 57 |
# endif |
| 58 |
# else |
| 59 |
/* this one was developed with sesse, |
| 60 |
* and is very similar to the one in snappy. |
| 61 |
* it does need a modern enough cpu with a fast multiplication. |
| 62 |
*/ |
| 63 |
# define IDX(h) (((h * 0x1e35a7bdU) >> (32 - HLOG - 8)) & (HSIZE - 1)) |
| 64 |
# endif |
| 65 |
#endif |
| 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 |
#if __GNUC__ >= 3 |
| 75 |
# define expect(expr,value) __builtin_expect ((expr),(value)) |
| 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) |
| 84 |
|
| 85 |
/* |
| 86 |
* compressed format |
| 87 |
* |
| 88 |
* 000LLLLL <L+1> ; literal, L+1=1..33 octets |
| 89 |
* LLLooooo oooooooo ; backref L+1=1..7 octets, o+1=1..4096 offset |
| 90 |
* 111ooooo LLLLLLLL oooooooo ; backref L+8 octets, o+1=1..4096 offset |
| 91 |
* |
| 92 |
*/ |
| 93 |
|
| 94 |
unsigned int |
| 95 |
lzf_compress (const void *const in_data, unsigned int in_len, |
| 96 |
void *out_data, unsigned int out_len |
| 97 |
#if LZF_STATE_ARG |
| 98 |
, LZF_STATE htab |
| 99 |
#endif |
| 100 |
) |
| 101 |
{ |
| 102 |
#if !LZF_STATE_ARG |
| 103 |
LZF_STATE htab; |
| 104 |
#endif |
| 105 |
const u8 *ip = (const u8 *)in_data; |
| 106 |
u8 *op = (u8 *)out_data; |
| 107 |
const u8 *in_end = ip + in_len; |
| 108 |
u8 *out_end = op + out_len; |
| 109 |
const u8 *ref; |
| 110 |
|
| 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 |
| 126 |
unsigned long off; |
| 127 |
#endif |
| 128 |
unsigned int hval; |
| 129 |
int lit; |
| 130 |
|
| 131 |
if (!in_len || !out_len) |
| 132 |
return 0; |
| 133 |
|
| 134 |
#if INIT_HTAB |
| 135 |
memset (htab, 0, sizeof (LZF_STATE)); |
| 136 |
#endif |
| 137 |
|
| 138 |
lit = 0; op++; /* start run */ |
| 139 |
|
| 140 |
hval = FRST (ip); |
| 141 |
while (ip < in_end - 2) |
| 142 |
{ |
| 143 |
LZF_HSLOT *hslot; |
| 144 |
|
| 145 |
hval = NEXT (hval, ip); |
| 146 |
hslot = htab + IDX (hval); |
| 147 |
ref = *hslot + LZF_HSLOT_BIAS; *hslot = ip - LZF_HSLOT_BIAS; |
| 148 |
|
| 149 |
if (1 |
| 150 |
#if INIT_HTAB |
| 151 |
&& ref < ip /* the next test will actually take care of this, but this is faster */ |
| 152 |
#endif |
| 153 |
&& (off = ip - ref - 1) < LZF_MAX_OFF |
| 154 |
&& ref > (u8 *)in_data |
| 155 |
&& ref[2] == ip[2] |
| 156 |
#if STRICT_ALIGN |
| 157 |
&& ((ref[1] << 8) | ref[0]) == ((ip[1] << 8) | ip[0]) |
| 158 |
#else |
| 159 |
&& *(u16 *)ref == *(u16 *)ip |
| 160 |
#endif |
| 161 |
) |
| 162 |
{ |
| 163 |
/* match found at *ref++ */ |
| 164 |
unsigned int len = 2; |
| 165 |
unsigned int maxlen = in_end - ip - len; |
| 166 |
maxlen = maxlen > LZF_MAX_REF ? LZF_MAX_REF : maxlen; |
| 167 |
|
| 168 |
if (expect_false (op + 3 + 1 >= out_end)) /* first a faster conservative test */ |
| 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++; |
| 202 |
while (len < maxlen && ref[len] == ip[len]); |
| 203 |
|
| 204 |
break; |
| 205 |
} |
| 206 |
|
| 207 |
len -= 2; /* len is now #octets - 1 */ |
| 208 |
ip++; |
| 209 |
|
| 210 |
if (len < 7) |
| 211 |
{ |
| 212 |
*op++ = (off >> 8) + (len << 5); |
| 213 |
} |
| 214 |
else |
| 215 |
{ |
| 216 |
*op++ = (off >> 8) + ( 7 << 5); |
| 217 |
*op++ = len - 7; |
| 218 |
} |
| 219 |
|
| 220 |
*op++ = off; |
| 221 |
|
| 222 |
lit = 0; op++; /* start run */ |
| 223 |
|
| 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 |
| 234 |
hval = FRST (ip); |
| 235 |
|
| 236 |
hval = NEXT (hval, ip); |
| 237 |
htab[IDX (hval)] = ip - LZF_HSLOT_BIAS; |
| 238 |
ip++; |
| 239 |
|
| 240 |
# if VERY_FAST && !ULTRA_FAST |
| 241 |
hval = NEXT (hval, ip); |
| 242 |
htab[IDX (hval)] = ip - LZF_HSLOT_BIAS; |
| 243 |
ip++; |
| 244 |
# endif |
| 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 |
} |
| 257 |
else |
| 258 |
{ |
| 259 |
/* one more literal byte we must copy */ |
| 260 |
if (expect_false (op >= out_end)) |
| 261 |
return 0; |
| 262 |
|
| 263 |
lit++; *op++ = *ip++; |
| 264 |
|
| 265 |
if (expect_false (lit == LZF_MAX_LIT)) |
| 266 |
{ |
| 267 |
op [- lit - 1] = lit - 1; /* stop run */ |
| 268 |
lit = 0; op++; /* start run */ |
| 269 |
} |
| 270 |
} |
| 271 |
} |
| 272 |
|
| 273 |
if (op + 3 > out_end) /* at most 3 bytes can be missing here */ |
| 274 |
return 0; |
| 275 |
|
| 276 |
while (ip < in_end) |
| 277 |
{ |
| 278 |
lit++; *op++ = *ip++; |
| 279 |
|
| 280 |
if (expect_false (lit == LZF_MAX_LIT)) |
| 281 |
{ |
| 282 |
op [- lit - 1] = lit - 1; /* stop run */ |
| 283 |
lit = 0; op++; /* start run */ |
| 284 |
} |
| 285 |
} |
| 286 |
|
| 287 |
op [- lit - 1] = lit - 1; /* end run */ |
| 288 |
op -= !lit; /* undo run if length is zero */ |
| 289 |
|
| 290 |
return op - (u8 *)out_data; |
| 291 |
} |
| 292 |
|