|
1 /* |
|
2 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. |
|
3 * |
|
4 * This code is free software; you can redistribute it and/or modify it |
|
5 * under the terms of the GNU General Public License version 2 only, as |
|
6 * published by the Free Software Foundation. Oracle designates this |
|
7 * particular file as subject to the "Classpath" exception as provided |
|
8 * by Oracle in the LICENSE file that accompanied this code. |
|
9 * |
|
10 * This code is distributed in the hope that it will be useful, but WITHOUT |
|
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
|
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
|
13 * version 2 for more details (a copy is included in the LICENSE file that |
|
14 * accompanied this code). |
|
15 * |
|
16 * You should have received a copy of the GNU General Public License version |
|
17 * 2 along with this work; if not, write to the Free Software Foundation, |
|
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. |
|
19 * |
|
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
|
21 * or visit www.oracle.com if you need additional information or have any |
|
22 * questions. |
|
23 */ |
|
24 |
|
25 /* inffast.c -- fast decoding |
|
26 * Copyright (C) 1995-2008, 2010 Mark Adler |
|
27 * For conditions of distribution and use, see copyright notice in zlib.h |
|
28 */ |
|
29 |
|
30 #include "zutil.h" |
|
31 #include "inftrees.h" |
|
32 #include "inflate.h" |
|
33 #include "inffast.h" |
|
34 |
|
35 #ifndef ASMINF |
|
36 |
|
37 /* Allow machine dependent optimization for post-increment or pre-increment. |
|
38 Based on testing to date, |
|
39 Pre-increment preferred for: |
|
40 - PowerPC G3 (Adler) |
|
41 - MIPS R5000 (Randers-Pehrson) |
|
42 Post-increment preferred for: |
|
43 - none |
|
44 No measurable difference: |
|
45 - Pentium III (Anderson) |
|
46 - M68060 (Nikl) |
|
47 */ |
|
48 #ifdef POSTINC |
|
49 # define OFF 0 |
|
50 # define PUP(a) *(a)++ |
|
51 #else |
|
52 # define OFF 1 |
|
53 # define PUP(a) *++(a) |
|
54 #endif |
|
55 |
|
56 /* |
|
57 Decode literal, length, and distance codes and write out the resulting |
|
58 literal and match bytes until either not enough input or output is |
|
59 available, an end-of-block is encountered, or a data error is encountered. |
|
60 When large enough input and output buffers are supplied to inflate(), for |
|
61 example, a 16K input buffer and a 64K output buffer, more than 95% of the |
|
62 inflate execution time is spent in this routine. |
|
63 |
|
64 Entry assumptions: |
|
65 |
|
66 state->mode == LEN |
|
67 strm->avail_in >= 6 |
|
68 strm->avail_out >= 258 |
|
69 start >= strm->avail_out |
|
70 state->bits < 8 |
|
71 |
|
72 On return, state->mode is one of: |
|
73 |
|
74 LEN -- ran out of enough output space or enough available input |
|
75 TYPE -- reached end of block code, inflate() to interpret next block |
|
76 BAD -- error in block data |
|
77 |
|
78 Notes: |
|
79 |
|
80 - The maximum input bits used by a length/distance pair is 15 bits for the |
|
81 length code, 5 bits for the length extra, 15 bits for the distance code, |
|
82 and 13 bits for the distance extra. This totals 48 bits, or six bytes. |
|
83 Therefore if strm->avail_in >= 6, then there is enough input to avoid |
|
84 checking for available input while decoding. |
|
85 |
|
86 - The maximum bytes that a single length/distance pair can output is 258 |
|
87 bytes, which is the maximum length that can be coded. inflate_fast() |
|
88 requires strm->avail_out >= 258 for each loop to avoid checking for |
|
89 output space. |
|
90 */ |
|
91 void ZLIB_INTERNAL inflate_fast(strm, start) |
|
92 z_streamp strm; |
|
93 unsigned start; /* inflate()'s starting value for strm->avail_out */ |
|
94 { |
|
95 struct inflate_state FAR *state; |
|
96 unsigned char FAR *in; /* local strm->next_in */ |
|
97 unsigned char FAR *last; /* while in < last, enough input available */ |
|
98 unsigned char FAR *out; /* local strm->next_out */ |
|
99 unsigned char FAR *beg; /* inflate()'s initial strm->next_out */ |
|
100 unsigned char FAR *end; /* while out < end, enough space available */ |
|
101 #ifdef INFLATE_STRICT |
|
102 unsigned dmax; /* maximum distance from zlib header */ |
|
103 #endif |
|
104 unsigned wsize; /* window size or zero if not using window */ |
|
105 unsigned whave; /* valid bytes in the window */ |
|
106 unsigned wnext; /* window write index */ |
|
107 unsigned char FAR *window; /* allocated sliding window, if wsize != 0 */ |
|
108 unsigned long hold; /* local strm->hold */ |
|
109 unsigned bits; /* local strm->bits */ |
|
110 code const FAR *lcode; /* local strm->lencode */ |
|
111 code const FAR *dcode; /* local strm->distcode */ |
|
112 unsigned lmask; /* mask for first level of length codes */ |
|
113 unsigned dmask; /* mask for first level of distance codes */ |
|
114 code here; /* retrieved table entry */ |
|
115 unsigned op; /* code bits, operation, extra bits, or */ |
|
116 /* window position, window bytes to copy */ |
|
117 unsigned len; /* match length, unused bytes */ |
|
118 unsigned dist; /* match distance */ |
|
119 unsigned char FAR *from; /* where to copy match from */ |
|
120 |
|
121 /* copy state to local variables */ |
|
122 state = (struct inflate_state FAR *)strm->state; |
|
123 in = strm->next_in - OFF; |
|
124 last = in + (strm->avail_in - 5); |
|
125 out = strm->next_out - OFF; |
|
126 beg = out - (start - strm->avail_out); |
|
127 end = out + (strm->avail_out - 257); |
|
128 #ifdef INFLATE_STRICT |
|
129 dmax = state->dmax; |
|
130 #endif |
|
131 wsize = state->wsize; |
|
132 whave = state->whave; |
|
133 wnext = state->wnext; |
|
134 window = state->window; |
|
135 hold = state->hold; |
|
136 bits = state->bits; |
|
137 lcode = state->lencode; |
|
138 dcode = state->distcode; |
|
139 lmask = (1U << state->lenbits) - 1; |
|
140 dmask = (1U << state->distbits) - 1; |
|
141 |
|
142 /* decode literals and length/distances until end-of-block or not enough |
|
143 input data or output space */ |
|
144 do { |
|
145 if (bits < 15) { |
|
146 hold += (unsigned long)(PUP(in)) << bits; |
|
147 bits += 8; |
|
148 hold += (unsigned long)(PUP(in)) << bits; |
|
149 bits += 8; |
|
150 } |
|
151 here = lcode[hold & lmask]; |
|
152 dolen: |
|
153 op = (unsigned)(here.bits); |
|
154 hold >>= op; |
|
155 bits -= op; |
|
156 op = (unsigned)(here.op); |
|
157 if (op == 0) { /* literal */ |
|
158 Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ? |
|
159 "inflate: literal '%c'\n" : |
|
160 "inflate: literal 0x%02x\n", here.val)); |
|
161 PUP(out) = (unsigned char)(here.val); |
|
162 } |
|
163 else if (op & 16) { /* length base */ |
|
164 len = (unsigned)(here.val); |
|
165 op &= 15; /* number of extra bits */ |
|
166 if (op) { |
|
167 if (bits < op) { |
|
168 hold += (unsigned long)(PUP(in)) << bits; |
|
169 bits += 8; |
|
170 } |
|
171 len += (unsigned)hold & ((1U << op) - 1); |
|
172 hold >>= op; |
|
173 bits -= op; |
|
174 } |
|
175 Tracevv((stderr, "inflate: length %u\n", len)); |
|
176 if (bits < 15) { |
|
177 hold += (unsigned long)(PUP(in)) << bits; |
|
178 bits += 8; |
|
179 hold += (unsigned long)(PUP(in)) << bits; |
|
180 bits += 8; |
|
181 } |
|
182 here = dcode[hold & dmask]; |
|
183 dodist: |
|
184 op = (unsigned)(here.bits); |
|
185 hold >>= op; |
|
186 bits -= op; |
|
187 op = (unsigned)(here.op); |
|
188 if (op & 16) { /* distance base */ |
|
189 dist = (unsigned)(here.val); |
|
190 op &= 15; /* number of extra bits */ |
|
191 if (bits < op) { |
|
192 hold += (unsigned long)(PUP(in)) << bits; |
|
193 bits += 8; |
|
194 if (bits < op) { |
|
195 hold += (unsigned long)(PUP(in)) << bits; |
|
196 bits += 8; |
|
197 } |
|
198 } |
|
199 dist += (unsigned)hold & ((1U << op) - 1); |
|
200 #ifdef INFLATE_STRICT |
|
201 if (dist > dmax) { |
|
202 strm->msg = (char *)"invalid distance too far back"; |
|
203 state->mode = BAD; |
|
204 break; |
|
205 } |
|
206 #endif |
|
207 hold >>= op; |
|
208 bits -= op; |
|
209 Tracevv((stderr, "inflate: distance %u\n", dist)); |
|
210 op = (unsigned)(out - beg); /* max distance in output */ |
|
211 if (dist > op) { /* see if copy from window */ |
|
212 op = dist - op; /* distance back in window */ |
|
213 if (op > whave) { |
|
214 if (state->sane) { |
|
215 strm->msg = |
|
216 (char *)"invalid distance too far back"; |
|
217 state->mode = BAD; |
|
218 break; |
|
219 } |
|
220 #ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR |
|
221 if (len <= op - whave) { |
|
222 do { |
|
223 PUP(out) = 0; |
|
224 } while (--len); |
|
225 continue; |
|
226 } |
|
227 len -= op - whave; |
|
228 do { |
|
229 PUP(out) = 0; |
|
230 } while (--op > whave); |
|
231 if (op == 0) { |
|
232 from = out - dist; |
|
233 do { |
|
234 PUP(out) = PUP(from); |
|
235 } while (--len); |
|
236 continue; |
|
237 } |
|
238 #endif |
|
239 } |
|
240 from = window - OFF; |
|
241 if (wnext == 0) { /* very common case */ |
|
242 from += wsize - op; |
|
243 if (op < len) { /* some from window */ |
|
244 len -= op; |
|
245 do { |
|
246 PUP(out) = PUP(from); |
|
247 } while (--op); |
|
248 from = out - dist; /* rest from output */ |
|
249 } |
|
250 } |
|
251 else if (wnext < op) { /* wrap around window */ |
|
252 from += wsize + wnext - op; |
|
253 op -= wnext; |
|
254 if (op < len) { /* some from end of window */ |
|
255 len -= op; |
|
256 do { |
|
257 PUP(out) = PUP(from); |
|
258 } while (--op); |
|
259 from = window - OFF; |
|
260 if (wnext < len) { /* some from start of window */ |
|
261 op = wnext; |
|
262 len -= op; |
|
263 do { |
|
264 PUP(out) = PUP(from); |
|
265 } while (--op); |
|
266 from = out - dist; /* rest from output */ |
|
267 } |
|
268 } |
|
269 } |
|
270 else { /* contiguous in window */ |
|
271 from += wnext - op; |
|
272 if (op < len) { /* some from window */ |
|
273 len -= op; |
|
274 do { |
|
275 PUP(out) = PUP(from); |
|
276 } while (--op); |
|
277 from = out - dist; /* rest from output */ |
|
278 } |
|
279 } |
|
280 while (len > 2) { |
|
281 PUP(out) = PUP(from); |
|
282 PUP(out) = PUP(from); |
|
283 PUP(out) = PUP(from); |
|
284 len -= 3; |
|
285 } |
|
286 if (len) { |
|
287 PUP(out) = PUP(from); |
|
288 if (len > 1) |
|
289 PUP(out) = PUP(from); |
|
290 } |
|
291 } |
|
292 else { |
|
293 from = out - dist; /* copy direct from output */ |
|
294 do { /* minimum length is three */ |
|
295 PUP(out) = PUP(from); |
|
296 PUP(out) = PUP(from); |
|
297 PUP(out) = PUP(from); |
|
298 len -= 3; |
|
299 } while (len > 2); |
|
300 if (len) { |
|
301 PUP(out) = PUP(from); |
|
302 if (len > 1) |
|
303 PUP(out) = PUP(from); |
|
304 } |
|
305 } |
|
306 } |
|
307 else if ((op & 64) == 0) { /* 2nd level distance code */ |
|
308 here = dcode[here.val + (hold & ((1U << op) - 1))]; |
|
309 goto dodist; |
|
310 } |
|
311 else { |
|
312 strm->msg = (char *)"invalid distance code"; |
|
313 state->mode = BAD; |
|
314 break; |
|
315 } |
|
316 } |
|
317 else if ((op & 64) == 0) { /* 2nd level length code */ |
|
318 here = lcode[here.val + (hold & ((1U << op) - 1))]; |
|
319 goto dolen; |
|
320 } |
|
321 else if (op & 32) { /* end-of-block */ |
|
322 Tracevv((stderr, "inflate: end of block\n")); |
|
323 state->mode = TYPE; |
|
324 break; |
|
325 } |
|
326 else { |
|
327 strm->msg = (char *)"invalid literal/length code"; |
|
328 state->mode = BAD; |
|
329 break; |
|
330 } |
|
331 } while (in < last && out < end); |
|
332 |
|
333 /* return unused bytes (on entry, bits < 8, so in won't go too far back) */ |
|
334 len = bits >> 3; |
|
335 in -= len; |
|
336 bits -= len << 3; |
|
337 hold &= (1U << bits) - 1; |
|
338 |
|
339 /* update state and return */ |
|
340 strm->next_in = in + OFF; |
|
341 strm->next_out = out + OFF; |
|
342 strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last)); |
|
343 strm->avail_out = (unsigned)(out < end ? |
|
344 257 + (end - out) : 257 - (out - end)); |
|
345 state->hold = hold; |
|
346 state->bits = bits; |
|
347 return; |
|
348 } |
|
349 |
|
350 /* |
|
351 inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe): |
|
352 - Using bit fields for code structure |
|
353 - Different op definition to avoid & for extra bits (do & for table bits) |
|
354 - Three separate decoding do-loops for direct, window, and wnext == 0 |
|
355 - Special case for distance > 1 copies to do overlapped load and store copy |
|
356 - Explicit branch predictions (based on measured branch probabilities) |
|
357 - Deferring match copy and interspersed it with decoding subsequent codes |
|
358 - Swapping literal/length else |
|
359 - Swapping window/direct else |
|
360 - Larger unrolled copy loops (three is about right) |
|
361 - Moving len -= 3 statement into middle of loop |
|
362 */ |
|
363 |
|
364 #endif /* !ASMINF */ |