1 | /* lzo1x_9x.c -- implementation of the LZO1X-999 compression algorithm |
---|
2 | |
---|
3 | This file is part of the LZO real-time data compression library. |
---|
4 | |
---|
5 | Copyright (C) 1996-2014 Markus Franz Xaver Johannes Oberhumer |
---|
6 | All Rights Reserved. |
---|
7 | |
---|
8 | The LZO library is free software; you can redistribute it and/or |
---|
9 | modify it under the terms of the GNU General Public License as |
---|
10 | published by the Free Software Foundation; either version 2 of |
---|
11 | the License, or (at your option) any later version. |
---|
12 | |
---|
13 | The LZO library is distributed in the hope that it will be useful, |
---|
14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
16 | GNU General Public License for more details. |
---|
17 | |
---|
18 | You should have received a copy of the GNU General Public License |
---|
19 | along with the LZO library; see the file COPYING. |
---|
20 | If not, write to the Free Software Foundation, Inc., |
---|
21 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. |
---|
22 | |
---|
23 | Markus F.X.J. Oberhumer |
---|
24 | <markus@oberhumer.com> |
---|
25 | http://www.oberhumer.com/opensource/lzo/ |
---|
26 | */ |
---|
27 | |
---|
28 | |
---|
29 | #if !defined(LZO1X) && !defined(LZO1Y) && !defined(LZO1Z) |
---|
30 | # define LZO1X 1 |
---|
31 | #endif |
---|
32 | |
---|
33 | #if defined(LZO1X) |
---|
34 | # include "config1x.h" |
---|
35 | #elif defined(LZO1Y) |
---|
36 | # include "config1y.h" |
---|
37 | #elif defined(LZO1Z) |
---|
38 | # include "config1z.h" |
---|
39 | #else |
---|
40 | # error |
---|
41 | #endif |
---|
42 | |
---|
43 | |
---|
44 | /*********************************************************************** |
---|
45 | // |
---|
46 | ************************************************************************/ |
---|
47 | |
---|
48 | #define SWD_N M4_MAX_OFFSET /* size of ring buffer */ |
---|
49 | #define SWD_THRESHOLD 1 /* lower limit for match length */ |
---|
50 | #define SWD_F 2048 /* upper limit for match length */ |
---|
51 | |
---|
52 | #define SWD_BEST_OFF (LZO_MAX3( M2_MAX_LEN, M3_MAX_LEN, M4_MAX_LEN ) + 1) |
---|
53 | |
---|
54 | #if defined(LZO1X) |
---|
55 | # define LZO_COMPRESS_T lzo1x_999_t |
---|
56 | # define lzo_swd_t lzo1x_999_swd_t |
---|
57 | #elif defined(LZO1Y) |
---|
58 | # define LZO_COMPRESS_T lzo1y_999_t |
---|
59 | # define lzo_swd_t lzo1y_999_swd_t |
---|
60 | # define lzo1x_999_compress_internal lzo1y_999_compress_internal |
---|
61 | # define lzo1x_999_compress_dict lzo1y_999_compress_dict |
---|
62 | # define lzo1x_999_compress_level lzo1y_999_compress_level |
---|
63 | # define lzo1x_999_compress lzo1y_999_compress |
---|
64 | #elif defined(LZO1Z) |
---|
65 | # define LZO_COMPRESS_T lzo1z_999_t |
---|
66 | # define lzo_swd_t lzo1z_999_swd_t |
---|
67 | # define lzo1x_999_compress_internal lzo1z_999_compress_internal |
---|
68 | # define lzo1x_999_compress_dict lzo1z_999_compress_dict |
---|
69 | # define lzo1x_999_compress_level lzo1z_999_compress_level |
---|
70 | # define lzo1x_999_compress lzo1z_999_compress |
---|
71 | #else |
---|
72 | # error |
---|
73 | #endif |
---|
74 | |
---|
75 | #if 0 |
---|
76 | # define HEAD3(b,p) \ |
---|
77 | ((((((lzo_xint)b[p]<<3)^b[p+1])<<3)^b[p+2]) & (SWD_HSIZE-1)) |
---|
78 | #endif |
---|
79 | #if 0 && (LZO_OPT_UNALIGNED32) && (LZO_ABI_LITTLE_ENDIAN) |
---|
80 | # define HEAD3(b,p) \ |
---|
81 | (((* (lzo_uint32_tp) &b[p]) ^ ((* (lzo_uint32_tp) &b[p])>>10)) & (SWD_HSIZE-1)) |
---|
82 | #endif |
---|
83 | |
---|
84 | #include "lzo_mchw.ch" |
---|
85 | |
---|
86 | |
---|
87 | /* this is a public functions, but there is no prototype in a header file */ |
---|
88 | LZO_EXTERN(int) |
---|
89 | lzo1x_999_compress_internal ( const lzo_bytep in , lzo_uint in_len, |
---|
90 | lzo_bytep out, lzo_uintp out_len, |
---|
91 | lzo_voidp wrkmem, |
---|
92 | const lzo_bytep dict, lzo_uint dict_len, |
---|
93 | lzo_callback_p cb, |
---|
94 | int try_lazy_parm, |
---|
95 | lzo_uint good_length, |
---|
96 | lzo_uint max_lazy, |
---|
97 | lzo_uint nice_length, |
---|
98 | lzo_uint max_chain, |
---|
99 | lzo_uint32_t flags ); |
---|
100 | |
---|
101 | |
---|
102 | /*********************************************************************** |
---|
103 | // |
---|
104 | ************************************************************************/ |
---|
105 | |
---|
106 | static lzo_bytep |
---|
107 | code_match ( LZO_COMPRESS_T *c, lzo_bytep op, lzo_uint m_len, lzo_uint m_off ) |
---|
108 | { |
---|
109 | lzo_uint x_len = m_len; |
---|
110 | lzo_uint x_off = m_off; |
---|
111 | |
---|
112 | c->match_bytes += (unsigned long) m_len; |
---|
113 | |
---|
114 | #if 0 |
---|
115 | /* |
---|
116 | static lzo_uint last_m_len = 0, last_m_off = 0; |
---|
117 | static lzo_uint prev_m_off[4]; |
---|
118 | static unsigned prev_m_off_ptr = 0; |
---|
119 | unsigned i; |
---|
120 | |
---|
121 | //if (m_len >= 3 && m_len <= M2_MAX_LEN && m_off <= M2_MAX_OFFSET) |
---|
122 | if (m_len >= 3 && m_len <= M2_MAX_LEN) |
---|
123 | { |
---|
124 | //if (m_len == last_m_len && m_off == last_m_off) |
---|
125 | //printf("last_m_len + last_m_off\n"); |
---|
126 | //else |
---|
127 | if (m_off == last_m_off) |
---|
128 | printf("last_m_off\n"); |
---|
129 | else |
---|
130 | { |
---|
131 | for (i = 0; i < 4; i++) |
---|
132 | if (m_off == prev_m_off[i]) |
---|
133 | printf("prev_m_off %u: %5ld\n",i,(long)m_off); |
---|
134 | } |
---|
135 | } |
---|
136 | last_m_len = m_len; |
---|
137 | last_m_off = prev_m_off[prev_m_off_ptr] = m_off; |
---|
138 | prev_m_off_ptr = (prev_m_off_ptr + 1) & 3; |
---|
139 | */ |
---|
140 | #endif |
---|
141 | |
---|
142 | assert(op > c->out); |
---|
143 | if (m_len == 2) |
---|
144 | { |
---|
145 | assert(m_off <= M1_MAX_OFFSET); |
---|
146 | assert(c->r1_lit > 0); assert(c->r1_lit < 4); |
---|
147 | m_off -= 1; |
---|
148 | #if defined(LZO1Z) |
---|
149 | *op++ = LZO_BYTE(M1_MARKER | (m_off >> 6)); |
---|
150 | *op++ = LZO_BYTE(m_off << 2); |
---|
151 | #else |
---|
152 | *op++ = LZO_BYTE(M1_MARKER | ((m_off & 3) << 2)); |
---|
153 | *op++ = LZO_BYTE(m_off >> 2); |
---|
154 | #endif |
---|
155 | c->m1a_m++; |
---|
156 | } |
---|
157 | #if defined(LZO1Z) |
---|
158 | else if (m_len <= M2_MAX_LEN && (m_off <= M2_MAX_OFFSET || m_off == c->last_m_off)) |
---|
159 | #else |
---|
160 | else if (m_len <= M2_MAX_LEN && m_off <= M2_MAX_OFFSET) |
---|
161 | #endif |
---|
162 | { |
---|
163 | assert(m_len >= 3); |
---|
164 | #if defined(LZO1X) |
---|
165 | m_off -= 1; |
---|
166 | *op++ = LZO_BYTE(((m_len - 1) << 5) | ((m_off & 7) << 2)); |
---|
167 | *op++ = LZO_BYTE(m_off >> 3); |
---|
168 | assert(op[-2] >= M2_MARKER); |
---|
169 | #elif defined(LZO1Y) |
---|
170 | m_off -= 1; |
---|
171 | *op++ = LZO_BYTE(((m_len + 1) << 4) | ((m_off & 3) << 2)); |
---|
172 | *op++ = LZO_BYTE(m_off >> 2); |
---|
173 | assert(op[-2] >= M2_MARKER); |
---|
174 | #elif defined(LZO1Z) |
---|
175 | if (m_off == c->last_m_off) |
---|
176 | *op++ = LZO_BYTE(((m_len - 1) << 5) | (0x700 >> 6)); |
---|
177 | else |
---|
178 | { |
---|
179 | m_off -= 1; |
---|
180 | *op++ = LZO_BYTE(((m_len - 1) << 5) | (m_off >> 6)); |
---|
181 | *op++ = LZO_BYTE(m_off << 2); |
---|
182 | } |
---|
183 | #endif |
---|
184 | c->m2_m++; |
---|
185 | } |
---|
186 | else if (m_len == M2_MIN_LEN && m_off <= MX_MAX_OFFSET && c->r1_lit >= 4) |
---|
187 | { |
---|
188 | assert(m_len == 3); |
---|
189 | assert(m_off > M2_MAX_OFFSET); |
---|
190 | m_off -= 1 + M2_MAX_OFFSET; |
---|
191 | #if defined(LZO1Z) |
---|
192 | *op++ = LZO_BYTE(M1_MARKER | (m_off >> 6)); |
---|
193 | *op++ = LZO_BYTE(m_off << 2); |
---|
194 | #else |
---|
195 | *op++ = LZO_BYTE(M1_MARKER | ((m_off & 3) << 2)); |
---|
196 | *op++ = LZO_BYTE(m_off >> 2); |
---|
197 | #endif |
---|
198 | c->m1b_m++; |
---|
199 | } |
---|
200 | else if (m_off <= M3_MAX_OFFSET) |
---|
201 | { |
---|
202 | assert(m_len >= 3); |
---|
203 | m_off -= 1; |
---|
204 | if (m_len <= M3_MAX_LEN) |
---|
205 | *op++ = LZO_BYTE(M3_MARKER | (m_len - 2)); |
---|
206 | else |
---|
207 | { |
---|
208 | m_len -= M3_MAX_LEN; |
---|
209 | *op++ = M3_MARKER | 0; |
---|
210 | while (m_len > 255) |
---|
211 | { |
---|
212 | m_len -= 255; |
---|
213 | *op++ = 0; |
---|
214 | } |
---|
215 | assert(m_len > 0); |
---|
216 | *op++ = LZO_BYTE(m_len); |
---|
217 | } |
---|
218 | #if defined(LZO1Z) |
---|
219 | *op++ = LZO_BYTE(m_off >> 6); |
---|
220 | *op++ = LZO_BYTE(m_off << 2); |
---|
221 | #else |
---|
222 | *op++ = LZO_BYTE(m_off << 2); |
---|
223 | *op++ = LZO_BYTE(m_off >> 6); |
---|
224 | #endif |
---|
225 | c->m3_m++; |
---|
226 | } |
---|
227 | else |
---|
228 | { |
---|
229 | lzo_uint k; |
---|
230 | |
---|
231 | assert(m_len >= 3); |
---|
232 | assert(m_off > 0x4000); assert(m_off <= 0xbfff); |
---|
233 | m_off -= 0x4000; |
---|
234 | k = (m_off & 0x4000) >> 11; |
---|
235 | if (m_len <= M4_MAX_LEN) |
---|
236 | *op++ = LZO_BYTE(M4_MARKER | k | (m_len - 2)); |
---|
237 | else |
---|
238 | { |
---|
239 | m_len -= M4_MAX_LEN; |
---|
240 | *op++ = LZO_BYTE(M4_MARKER | k | 0); |
---|
241 | while (m_len > 255) |
---|
242 | { |
---|
243 | m_len -= 255; |
---|
244 | *op++ = 0; |
---|
245 | } |
---|
246 | assert(m_len > 0); |
---|
247 | *op++ = LZO_BYTE(m_len); |
---|
248 | } |
---|
249 | #if defined(LZO1Z) |
---|
250 | *op++ = LZO_BYTE(m_off >> 6); |
---|
251 | *op++ = LZO_BYTE(m_off << 2); |
---|
252 | #else |
---|
253 | *op++ = LZO_BYTE(m_off << 2); |
---|
254 | *op++ = LZO_BYTE(m_off >> 6); |
---|
255 | #endif |
---|
256 | c->m4_m++; |
---|
257 | } |
---|
258 | |
---|
259 | c->last_m_len = x_len; |
---|
260 | c->last_m_off = x_off; |
---|
261 | return op; |
---|
262 | } |
---|
263 | |
---|
264 | |
---|
265 | static lzo_bytep |
---|
266 | STORE_RUN ( LZO_COMPRESS_T *c, lzo_bytep op, const lzo_bytep ii, lzo_uint t ) |
---|
267 | { |
---|
268 | c->lit_bytes += (unsigned long) t; |
---|
269 | |
---|
270 | if (op == c->out && t <= 238) |
---|
271 | { |
---|
272 | *op++ = LZO_BYTE(17 + t); |
---|
273 | } |
---|
274 | else if (t <= 3) |
---|
275 | { |
---|
276 | #if defined(LZO1Z) |
---|
277 | op[-1] = LZO_BYTE(op[-1] | t); |
---|
278 | #else |
---|
279 | op[-2] = LZO_BYTE(op[-2] | t); |
---|
280 | #endif |
---|
281 | c->lit1_r++; |
---|
282 | } |
---|
283 | else if (t <= 18) |
---|
284 | { |
---|
285 | *op++ = LZO_BYTE(t - 3); |
---|
286 | c->lit2_r++; |
---|
287 | } |
---|
288 | else |
---|
289 | { |
---|
290 | lzo_uint tt = t - 18; |
---|
291 | |
---|
292 | *op++ = 0; |
---|
293 | while (tt > 255) |
---|
294 | { |
---|
295 | tt -= 255; |
---|
296 | *op++ = 0; |
---|
297 | } |
---|
298 | assert(tt > 0); |
---|
299 | *op++ = LZO_BYTE(tt); |
---|
300 | c->lit3_r++; |
---|
301 | } |
---|
302 | do *op++ = *ii++; while (--t > 0); |
---|
303 | |
---|
304 | return op; |
---|
305 | } |
---|
306 | |
---|
307 | |
---|
308 | static lzo_bytep |
---|
309 | code_run ( LZO_COMPRESS_T *c, lzo_bytep op, const lzo_bytep ii, |
---|
310 | lzo_uint lit, lzo_uint m_len ) |
---|
311 | { |
---|
312 | if (lit > 0) |
---|
313 | { |
---|
314 | assert(m_len >= 2); |
---|
315 | op = STORE_RUN(c,op,ii,lit); |
---|
316 | c->r1_m_len = m_len; |
---|
317 | c->r1_lit = lit; |
---|
318 | } |
---|
319 | else |
---|
320 | { |
---|
321 | assert(m_len >= 3); |
---|
322 | c->r1_m_len = 0; |
---|
323 | c->r1_lit = 0; |
---|
324 | } |
---|
325 | |
---|
326 | return op; |
---|
327 | } |
---|
328 | |
---|
329 | |
---|
330 | /*********************************************************************** |
---|
331 | // |
---|
332 | ************************************************************************/ |
---|
333 | |
---|
334 | static lzo_uint |
---|
335 | len_of_coded_match ( lzo_uint m_len, lzo_uint m_off, lzo_uint lit ) |
---|
336 | { |
---|
337 | lzo_uint n = 4; |
---|
338 | |
---|
339 | if (m_len < 2) |
---|
340 | return 0; |
---|
341 | if (m_len == 2) |
---|
342 | return (m_off <= M1_MAX_OFFSET && lit > 0 && lit < 4) ? 2 : 0; |
---|
343 | if (m_len <= M2_MAX_LEN && m_off <= M2_MAX_OFFSET) |
---|
344 | return 2; |
---|
345 | if (m_len == M2_MIN_LEN && m_off <= MX_MAX_OFFSET && lit >= 4) |
---|
346 | return 2; |
---|
347 | if (m_off <= M3_MAX_OFFSET) |
---|
348 | { |
---|
349 | if (m_len <= M3_MAX_LEN) |
---|
350 | return 3; |
---|
351 | m_len -= M3_MAX_LEN; |
---|
352 | while (m_len > 255) |
---|
353 | { |
---|
354 | m_len -= 255; |
---|
355 | n++; |
---|
356 | } |
---|
357 | return n; |
---|
358 | } |
---|
359 | if (m_off <= M4_MAX_OFFSET) |
---|
360 | { |
---|
361 | if (m_len <= M4_MAX_LEN) |
---|
362 | return 3; |
---|
363 | m_len -= M4_MAX_LEN; |
---|
364 | while (m_len > 255) |
---|
365 | { |
---|
366 | m_len -= 255; |
---|
367 | n++; |
---|
368 | } |
---|
369 | return n; |
---|
370 | } |
---|
371 | return 0; |
---|
372 | } |
---|
373 | |
---|
374 | |
---|
375 | static lzo_uint |
---|
376 | min_gain(lzo_uint ahead, lzo_uint lit1, lzo_uint lit2, lzo_uint l1, lzo_uint l2, lzo_uint l3) |
---|
377 | { |
---|
378 | lzo_uint lazy_match_min_gain; |
---|
379 | |
---|
380 | assert (ahead >= 1); |
---|
381 | lazy_match_min_gain = ahead; |
---|
382 | |
---|
383 | #if 0 |
---|
384 | if (l3) |
---|
385 | lit2 -= ahead; |
---|
386 | #endif |
---|
387 | |
---|
388 | if (lit1 <= 3) |
---|
389 | lazy_match_min_gain += (lit2 <= 3) ? 0 : 2; |
---|
390 | else if (lit1 <= 18) |
---|
391 | lazy_match_min_gain += (lit2 <= 18) ? 0 : 1; |
---|
392 | |
---|
393 | lazy_match_min_gain += (l2 - l1) * 2; |
---|
394 | if (l3) |
---|
395 | lazy_match_min_gain -= (ahead - l3) * 2; |
---|
396 | |
---|
397 | if ((lzo_int) lazy_match_min_gain < 0) |
---|
398 | lazy_match_min_gain = 0; |
---|
399 | |
---|
400 | #if 0 |
---|
401 | if (l1 == 2) |
---|
402 | if (lazy_match_min_gain == 0) |
---|
403 | lazy_match_min_gain = 1; |
---|
404 | #endif |
---|
405 | |
---|
406 | return lazy_match_min_gain; |
---|
407 | } |
---|
408 | |
---|
409 | |
---|
410 | /*********************************************************************** |
---|
411 | // |
---|
412 | ************************************************************************/ |
---|
413 | |
---|
414 | #if !defined(NDEBUG) |
---|
415 | static |
---|
416 | void assert_match( const lzo_swd_p swd, lzo_uint m_len, lzo_uint m_off ) |
---|
417 | { |
---|
418 | const LZO_COMPRESS_T *c = swd->c; |
---|
419 | lzo_uint d_off; |
---|
420 | |
---|
421 | assert(m_len >= 2); |
---|
422 | if (m_off <= (lzo_uint) (c->bp - c->in)) |
---|
423 | { |
---|
424 | assert(c->bp - m_off + m_len < c->ip); |
---|
425 | assert(lzo_memcmp(c->bp, c->bp - m_off, m_len) == 0); |
---|
426 | } |
---|
427 | else |
---|
428 | { |
---|
429 | assert(swd->dict != NULL); |
---|
430 | d_off = m_off - (lzo_uint) (c->bp - c->in); |
---|
431 | assert(d_off <= swd->dict_len); |
---|
432 | if (m_len > d_off) |
---|
433 | { |
---|
434 | assert(lzo_memcmp(c->bp, swd->dict_end - d_off, d_off) == 0); |
---|
435 | assert(c->in + m_len - d_off < c->ip); |
---|
436 | assert(lzo_memcmp(c->bp + d_off, c->in, m_len - d_off) == 0); |
---|
437 | } |
---|
438 | else |
---|
439 | { |
---|
440 | assert(lzo_memcmp(c->bp, swd->dict_end - d_off, m_len) == 0); |
---|
441 | } |
---|
442 | } |
---|
443 | } |
---|
444 | #else |
---|
445 | # define assert_match(a,b,c) ((void)0) |
---|
446 | #endif |
---|
447 | |
---|
448 | |
---|
449 | #if defined(SWD_BEST_OFF) |
---|
450 | |
---|
451 | static void |
---|
452 | better_match ( const lzo_swd_p swd, lzo_uint *m_len, lzo_uint *m_off ) |
---|
453 | { |
---|
454 | #if defined(LZO1Z) |
---|
455 | const LZO_COMPRESS_T *c = swd->c; |
---|
456 | #endif |
---|
457 | |
---|
458 | if (*m_len <= M2_MIN_LEN) |
---|
459 | return; |
---|
460 | #if defined(LZO1Z) |
---|
461 | if (*m_off == c->last_m_off && *m_len <= M2_MAX_LEN) |
---|
462 | return; |
---|
463 | #if 1 |
---|
464 | if (*m_len >= M2_MIN_LEN + 1 && *m_len <= M2_MAX_LEN + 1 && |
---|
465 | c->last_m_off && swd->best_off[*m_len-1] == c->last_m_off) |
---|
466 | { |
---|
467 | *m_len = *m_len - 1; |
---|
468 | *m_off = swd->best_off[*m_len]; |
---|
469 | return; |
---|
470 | } |
---|
471 | #endif |
---|
472 | #endif |
---|
473 | |
---|
474 | if (*m_off <= M2_MAX_OFFSET) |
---|
475 | return; |
---|
476 | |
---|
477 | #if 1 |
---|
478 | /* M3/M4 -> M2 */ |
---|
479 | if (*m_off > M2_MAX_OFFSET && |
---|
480 | *m_len >= M2_MIN_LEN + 1 && *m_len <= M2_MAX_LEN + 1 && |
---|
481 | swd->best_off[*m_len-1] && swd->best_off[*m_len-1] <= M2_MAX_OFFSET) |
---|
482 | { |
---|
483 | *m_len = *m_len - 1; |
---|
484 | *m_off = swd->best_off[*m_len]; |
---|
485 | return; |
---|
486 | } |
---|
487 | #endif |
---|
488 | |
---|
489 | #if 1 |
---|
490 | /* M4 -> M2 */ |
---|
491 | if (*m_off > M3_MAX_OFFSET && |
---|
492 | *m_len >= M4_MAX_LEN + 1 && *m_len <= M2_MAX_LEN + 2 && |
---|
493 | swd->best_off[*m_len-2] && swd->best_off[*m_len-2] <= M2_MAX_OFFSET) |
---|
494 | { |
---|
495 | *m_len = *m_len - 2; |
---|
496 | *m_off = swd->best_off[*m_len]; |
---|
497 | return; |
---|
498 | } |
---|
499 | #endif |
---|
500 | |
---|
501 | #if 1 |
---|
502 | /* M4 -> M3 */ |
---|
503 | if (*m_off > M3_MAX_OFFSET && |
---|
504 | *m_len >= M4_MAX_LEN + 1 && *m_len <= M3_MAX_LEN + 1 && |
---|
505 | swd->best_off[*m_len-1] && swd->best_off[*m_len-1] <= M3_MAX_OFFSET) |
---|
506 | { |
---|
507 | *m_len = *m_len - 1; |
---|
508 | *m_off = swd->best_off[*m_len]; |
---|
509 | } |
---|
510 | #endif |
---|
511 | } |
---|
512 | |
---|
513 | #endif |
---|
514 | |
---|
515 | |
---|
516 | /*********************************************************************** |
---|
517 | // |
---|
518 | ************************************************************************/ |
---|
519 | |
---|
520 | LZO_PUBLIC(int) |
---|
521 | lzo1x_999_compress_internal ( const lzo_bytep in , lzo_uint in_len, |
---|
522 | lzo_bytep out, lzo_uintp out_len, |
---|
523 | lzo_voidp wrkmem, |
---|
524 | const lzo_bytep dict, lzo_uint dict_len, |
---|
525 | lzo_callback_p cb, |
---|
526 | int try_lazy_parm, |
---|
527 | lzo_uint good_length, |
---|
528 | lzo_uint max_lazy, |
---|
529 | lzo_uint nice_length, |
---|
530 | lzo_uint max_chain, |
---|
531 | lzo_uint32_t flags ) |
---|
532 | { |
---|
533 | lzo_bytep op; |
---|
534 | const lzo_bytep ii; |
---|
535 | lzo_uint lit; |
---|
536 | lzo_uint m_len, m_off; |
---|
537 | LZO_COMPRESS_T cc; |
---|
538 | LZO_COMPRESS_T * const c = &cc; |
---|
539 | lzo_swd_p const swd = (lzo_swd_p) wrkmem; |
---|
540 | lzo_uint try_lazy; |
---|
541 | int r; |
---|
542 | |
---|
543 | /* sanity check */ |
---|
544 | #if defined(LZO1X) |
---|
545 | LZO_COMPILE_TIME_ASSERT(LZO1X_999_MEM_COMPRESS >= SIZEOF_LZO_SWD_T) |
---|
546 | #elif defined(LZO1Y) |
---|
547 | LZO_COMPILE_TIME_ASSERT(LZO1Y_999_MEM_COMPRESS >= SIZEOF_LZO_SWD_T) |
---|
548 | #elif defined(LZO1Z) |
---|
549 | LZO_COMPILE_TIME_ASSERT(LZO1Z_999_MEM_COMPRESS >= SIZEOF_LZO_SWD_T) |
---|
550 | #else |
---|
551 | # error |
---|
552 | #endif |
---|
553 | |
---|
554 | /* setup parameter defaults */ |
---|
555 | /* number of lazy match tries */ |
---|
556 | try_lazy = (lzo_uint) try_lazy_parm; |
---|
557 | if (try_lazy_parm < 0) |
---|
558 | try_lazy = 1; |
---|
559 | /* reduce lazy match search if we already have a match with this length */ |
---|
560 | if (good_length == 0) |
---|
561 | good_length = 32; |
---|
562 | /* do not try a lazy match if we already have a match with this length */ |
---|
563 | if (max_lazy == 0) |
---|
564 | max_lazy = 32; |
---|
565 | /* stop searching for longer matches than this one */ |
---|
566 | if (nice_length == 0) |
---|
567 | nice_length = 0; |
---|
568 | /* don't search more positions than this */ |
---|
569 | if (max_chain == 0) |
---|
570 | max_chain = SWD_MAX_CHAIN; |
---|
571 | |
---|
572 | c->init = 0; |
---|
573 | c->ip = c->in = in; |
---|
574 | c->in_end = in + in_len; |
---|
575 | c->out = out; |
---|
576 | c->cb = cb; |
---|
577 | c->m1a_m = c->m1b_m = c->m2_m = c->m3_m = c->m4_m = 0; |
---|
578 | c->lit1_r = c->lit2_r = c->lit3_r = 0; |
---|
579 | |
---|
580 | op = out; |
---|
581 | ii = c->ip; /* point to start of literal run */ |
---|
582 | lit = 0; |
---|
583 | c->r1_lit = c->r1_m_len = 0; |
---|
584 | |
---|
585 | r = init_match(c,swd,dict,dict_len,flags); |
---|
586 | if (r != 0) |
---|
587 | return r; |
---|
588 | if (max_chain > 0) |
---|
589 | swd->max_chain = max_chain; |
---|
590 | if (nice_length > 0) |
---|
591 | swd->nice_length = nice_length; |
---|
592 | |
---|
593 | r = find_match(c,swd,0,0); |
---|
594 | if (r != 0) |
---|
595 | return r; |
---|
596 | while (c->look > 0) |
---|
597 | { |
---|
598 | lzo_uint ahead; |
---|
599 | lzo_uint max_ahead; |
---|
600 | lzo_uint l1, l2, l3; |
---|
601 | |
---|
602 | c->codesize = pd(op, out); |
---|
603 | |
---|
604 | m_len = c->m_len; |
---|
605 | m_off = c->m_off; |
---|
606 | |
---|
607 | assert(c->bp == c->ip - c->look); |
---|
608 | assert(c->bp >= in); |
---|
609 | if (lit == 0) |
---|
610 | ii = c->bp; |
---|
611 | assert(ii + lit == c->bp); |
---|
612 | assert(swd->b_char == *(c->bp)); |
---|
613 | |
---|
614 | if ( m_len < 2 || |
---|
615 | (m_len == 2 && (m_off > M1_MAX_OFFSET || lit == 0 || lit >= 4)) || |
---|
616 | #if 1 |
---|
617 | /* Do not accept this match for compressed-data compatibility |
---|
618 | * with LZO v1.01 and before |
---|
619 | * [ might be a problem for decompress() and optimize() ] |
---|
620 | */ |
---|
621 | (m_len == 2 && op == out) || |
---|
622 | #endif |
---|
623 | (op == out && lit == 0)) |
---|
624 | { |
---|
625 | /* a literal */ |
---|
626 | m_len = 0; |
---|
627 | } |
---|
628 | else if (m_len == M2_MIN_LEN) |
---|
629 | { |
---|
630 | /* compression ratio improves if we code a literal in some cases */ |
---|
631 | if (m_off > MX_MAX_OFFSET && lit >= 4) |
---|
632 | m_len = 0; |
---|
633 | } |
---|
634 | |
---|
635 | if (m_len == 0) |
---|
636 | { |
---|
637 | /* a literal */ |
---|
638 | lit++; |
---|
639 | swd->max_chain = max_chain; |
---|
640 | r = find_match(c,swd,1,0); |
---|
641 | assert(r == 0); LZO_UNUSED(r); |
---|
642 | continue; |
---|
643 | } |
---|
644 | |
---|
645 | /* a match */ |
---|
646 | #if defined(SWD_BEST_OFF) |
---|
647 | if (swd->use_best_off) |
---|
648 | better_match(swd,&m_len,&m_off); |
---|
649 | #endif |
---|
650 | assert_match(swd,m_len,m_off); |
---|
651 | |
---|
652 | |
---|
653 | /* shall we try a lazy match ? */ |
---|
654 | ahead = 0; |
---|
655 | if (try_lazy == 0 || m_len >= max_lazy) |
---|
656 | { |
---|
657 | /* no */ |
---|
658 | l1 = 0; |
---|
659 | max_ahead = 0; |
---|
660 | } |
---|
661 | else |
---|
662 | { |
---|
663 | /* yes, try a lazy match */ |
---|
664 | l1 = len_of_coded_match(m_len,m_off,lit); |
---|
665 | assert(l1 > 0); |
---|
666 | #if 1 |
---|
667 | max_ahead = LZO_MIN(try_lazy, l1 - 1); |
---|
668 | #else |
---|
669 | max_ahead = LZO_MIN3(try_lazy, l1, m_len - 1); |
---|
670 | #endif |
---|
671 | } |
---|
672 | |
---|
673 | |
---|
674 | while (ahead < max_ahead && c->look > m_len) |
---|
675 | { |
---|
676 | lzo_uint lazy_match_min_gain; |
---|
677 | |
---|
678 | if (m_len >= good_length) |
---|
679 | swd->max_chain = max_chain >> 2; |
---|
680 | else |
---|
681 | swd->max_chain = max_chain; |
---|
682 | r = find_match(c,swd,1,0); |
---|
683 | ahead++; |
---|
684 | |
---|
685 | assert(r == 0); LZO_UNUSED(r); |
---|
686 | assert(c->look > 0); |
---|
687 | assert(ii + lit + ahead == c->bp); |
---|
688 | |
---|
689 | #if defined(LZO1Z) |
---|
690 | if (m_off == c->last_m_off && c->m_off != c->last_m_off) |
---|
691 | if (m_len >= M2_MIN_LEN && m_len <= M2_MAX_LEN) |
---|
692 | c->m_len = 0; |
---|
693 | #endif |
---|
694 | if (c->m_len < m_len) |
---|
695 | continue; |
---|
696 | #if 1 |
---|
697 | if (c->m_len == m_len && c->m_off >= m_off) |
---|
698 | continue; |
---|
699 | #endif |
---|
700 | #if defined(SWD_BEST_OFF) |
---|
701 | if (swd->use_best_off) |
---|
702 | better_match(swd,&c->m_len,&c->m_off); |
---|
703 | #endif |
---|
704 | l2 = len_of_coded_match(c->m_len,c->m_off,lit+ahead); |
---|
705 | if (l2 == 0) |
---|
706 | continue; |
---|
707 | #if 0 |
---|
708 | if (c->m_len == m_len && l2 >= l1) |
---|
709 | continue; |
---|
710 | #endif |
---|
711 | |
---|
712 | |
---|
713 | #if 1 |
---|
714 | /* compressed-data compatibility [see above] */ |
---|
715 | l3 = (op == out) ? 0 : len_of_coded_match(ahead,m_off,lit); |
---|
716 | #else |
---|
717 | l3 = len_of_coded_match(ahead,m_off,lit); |
---|
718 | #endif |
---|
719 | |
---|
720 | lazy_match_min_gain = min_gain(ahead,lit,lit+ahead,l1,l2,l3); |
---|
721 | if (c->m_len >= m_len + lazy_match_min_gain) |
---|
722 | { |
---|
723 | c->lazy++; |
---|
724 | assert_match(swd,c->m_len,c->m_off); |
---|
725 | |
---|
726 | if (l3) |
---|
727 | { |
---|
728 | /* code previous run */ |
---|
729 | op = code_run(c,op,ii,lit,ahead); |
---|
730 | lit = 0; |
---|
731 | /* code shortened match */ |
---|
732 | op = code_match(c,op,ahead,m_off); |
---|
733 | } |
---|
734 | else |
---|
735 | { |
---|
736 | lit += ahead; |
---|
737 | assert(ii + lit == c->bp); |
---|
738 | } |
---|
739 | goto lazy_match_done; |
---|
740 | } |
---|
741 | } |
---|
742 | |
---|
743 | |
---|
744 | assert(ii + lit + ahead == c->bp); |
---|
745 | |
---|
746 | /* 1 - code run */ |
---|
747 | op = code_run(c,op,ii,lit,m_len); |
---|
748 | lit = 0; |
---|
749 | |
---|
750 | /* 2 - code match */ |
---|
751 | op = code_match(c,op,m_len,m_off); |
---|
752 | swd->max_chain = max_chain; |
---|
753 | r = find_match(c,swd,m_len,1+ahead); |
---|
754 | assert(r == 0); LZO_UNUSED(r); |
---|
755 | |
---|
756 | lazy_match_done: ; |
---|
757 | } |
---|
758 | |
---|
759 | |
---|
760 | /* store final run */ |
---|
761 | if (lit > 0) |
---|
762 | op = STORE_RUN(c,op,ii,lit); |
---|
763 | |
---|
764 | #if defined(LZO_EOF_CODE) |
---|
765 | *op++ = M4_MARKER | 1; |
---|
766 | *op++ = 0; |
---|
767 | *op++ = 0; |
---|
768 | #endif |
---|
769 | |
---|
770 | c->codesize = pd(op, out); |
---|
771 | assert(c->textsize == in_len); |
---|
772 | |
---|
773 | *out_len = pd(op, out); |
---|
774 | |
---|
775 | if (c->cb && c->cb->nprogress) |
---|
776 | (*c->cb->nprogress)(c->cb, c->textsize, c->codesize, 0); |
---|
777 | |
---|
778 | #if 0 |
---|
779 | printf("%ld %ld -> %ld %ld: %ld %ld %ld %ld %ld %ld: %ld %ld %ld %ld\n", |
---|
780 | (long) c->textsize, (long) in_len, (long) c->codesize, |
---|
781 | c->match_bytes, c->m1a_m, c->m1b_m, c->m2_m, c->m3_m, c->m4_m, |
---|
782 | c->lit_bytes, c->lit1_r, c->lit2_r, c->lit3_r, c->lazy); |
---|
783 | #endif |
---|
784 | assert(c->lit_bytes + c->match_bytes == in_len); |
---|
785 | |
---|
786 | return LZO_E_OK; |
---|
787 | } |
---|
788 | |
---|
789 | |
---|
790 | /*********************************************************************** |
---|
791 | // |
---|
792 | ************************************************************************/ |
---|
793 | |
---|
794 | LZO_PUBLIC(int) |
---|
795 | lzo1x_999_compress_level ( const lzo_bytep in , lzo_uint in_len, |
---|
796 | lzo_bytep out, lzo_uintp out_len, |
---|
797 | lzo_voidp wrkmem, |
---|
798 | const lzo_bytep dict, lzo_uint dict_len, |
---|
799 | lzo_callback_p cb, |
---|
800 | int compression_level ) |
---|
801 | { |
---|
802 | static const struct |
---|
803 | { |
---|
804 | int try_lazy_parm; |
---|
805 | lzo_uint good_length; |
---|
806 | lzo_uint max_lazy; |
---|
807 | lzo_uint nice_length; |
---|
808 | lzo_uint max_chain; |
---|
809 | lzo_uint32_t flags; |
---|
810 | } c[9] = { |
---|
811 | /* faster compression */ |
---|
812 | { 0, 0, 0, 8, 4, 0 }, |
---|
813 | { 0, 0, 0, 16, 8, 0 }, |
---|
814 | { 0, 0, 0, 32, 16, 0 }, |
---|
815 | { 1, 4, 4, 16, 16, 0 }, |
---|
816 | { 1, 8, 16, 32, 32, 0 }, |
---|
817 | { 1, 8, 16, 128, 128, 0 }, |
---|
818 | { 2, 8, 32, 128, 256, 0 }, |
---|
819 | { 2, 32, 128, SWD_F, 2048, 1 }, |
---|
820 | { 2, SWD_F, SWD_F, SWD_F, 4096, 1 } |
---|
821 | /* max. compression */ |
---|
822 | }; |
---|
823 | |
---|
824 | if (compression_level < 1 || compression_level > 9) |
---|
825 | return LZO_E_ERROR; |
---|
826 | |
---|
827 | compression_level -= 1; |
---|
828 | return lzo1x_999_compress_internal(in, in_len, out, out_len, wrkmem, |
---|
829 | dict, dict_len, cb, |
---|
830 | c[compression_level].try_lazy_parm, |
---|
831 | c[compression_level].good_length, |
---|
832 | c[compression_level].max_lazy, |
---|
833 | #if 0 |
---|
834 | c[compression_level].nice_length, |
---|
835 | #else |
---|
836 | 0, |
---|
837 | #endif |
---|
838 | c[compression_level].max_chain, |
---|
839 | c[compression_level].flags); |
---|
840 | } |
---|
841 | |
---|
842 | |
---|
843 | /*********************************************************************** |
---|
844 | // |
---|
845 | ************************************************************************/ |
---|
846 | |
---|
847 | LZO_PUBLIC(int) |
---|
848 | lzo1x_999_compress_dict ( const lzo_bytep in , lzo_uint in_len, |
---|
849 | lzo_bytep out, lzo_uintp out_len, |
---|
850 | lzo_voidp wrkmem, |
---|
851 | const lzo_bytep dict, lzo_uint dict_len ) |
---|
852 | { |
---|
853 | return lzo1x_999_compress_level(in, in_len, out, out_len, wrkmem, |
---|
854 | dict, dict_len, 0, 8); |
---|
855 | } |
---|
856 | |
---|
857 | LZO_PUBLIC(int) |
---|
858 | lzo1x_999_compress ( const lzo_bytep in , lzo_uint in_len, |
---|
859 | lzo_bytep out, lzo_uintp out_len, |
---|
860 | lzo_voidp wrkmem ) |
---|
861 | { |
---|
862 | return lzo1x_999_compress_level(in, in_len, out, out_len, wrkmem, |
---|
863 | NULL, 0, (lzo_callback_p) 0, 8); |
---|
864 | } |
---|
865 | |
---|
866 | |
---|
867 | /* |
---|
868 | vi:ts=4:et |
---|
869 | */ |
---|
870 | |
---|