1/* 2 * Copyright (c) 2008 Apple Inc. All rights reserved. 3 * 4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ 5 * 6 * This file contains Original Code and/or Modifications of Original Code 7 * as defined in and that are subject to the Apple Public Source License 8 * Version 2.0 (the 'License'). You may not use this file except in 9 * compliance with the License. The rights granted to you under the License 10 * may not be used to create, or enable the creation or redistribution of, 11 * unlawful or unlicensed copies of an Apple operating system, or to 12 * circumvent, violate, or enable the circumvention or violation of, any 13 * terms of an Apple operating system software license agreement. 14 * 15 * Please obtain a copy of the License at 16 * http://www.opensource.apple.com/apsl/ and read it before using this file. 17 * 18 * The Original Code and all software distributed under the License are 19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER 20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, 21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, 22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. 23 * Please see the License for the specific language governing rights and 24 * limitations under the License. 25 * 26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ 27 */ 28/* inffast.c -- fast decoding 29 * Copyright (C) 1995-2004 Mark Adler 30 * For conditions of distribution and use, see copyright notice in zlib.h 31 */ 32 33 34#include "zutil.h" 35#include "inftrees.h" 36#include "inflate.h" 37#include "inffast.h" 38 39#ifndef ASMINF 40 41/* Allow machine dependent optimization for post-increment or pre-increment. 42 Based on testing to date, 43 Pre-increment preferred for: 44 - PowerPC G3 (Adler) 45 - MIPS R5000 (Randers-Pehrson) 46 Post-increment preferred for: 47 - none 48 No measurable difference: 49 - Pentium III (Anderson) 50 - M68060 (Nikl) 51 */ 52#ifdef POSTINC 53# define OFF 0 54# define PUP(a) *(a)++ 55#else 56# define OFF 1 57# define PUP(a) *++(a) 58#endif 59 60/* 61 Decode literal, length, and distance codes and write out the resulting 62 literal and match bytes until either not enough input or output is 63 available, an end-of-block is encountered, or a data error is encountered. 64 When large enough input and output buffers are supplied to inflate(), for 65 example, a 16K input buffer and a 64K output buffer, more than 95% of the 66 inflate execution time is spent in this routine. 67 68 Entry assumptions: 69 70 state->mode == LEN 71 strm->avail_in >= 6 72 strm->avail_out >= 258 73 start >= strm->avail_out 74 state->bits < 8 75 76 On return, state->mode is one of: 77 78 LEN -- ran out of enough output space or enough available input 79 TYPE -- reached end of block code, inflate() to interpret next block 80 BAD -- error in block data 81 82 Notes: 83 84 - The maximum input bits used by a length/distance pair is 15 bits for the 85 length code, 5 bits for the length extra, 15 bits for the distance code, 86 and 13 bits for the distance extra. This totals 48 bits, or six bytes. 87 Therefore if strm->avail_in >= 6, then there is enough input to avoid 88 checking for available input while decoding. 89 90 - The maximum bytes that a single length/distance pair can output is 258 91 bytes, which is the maximum length that can be coded. inflate_fast() 92 requires strm->avail_out >= 258 for each loop to avoid checking for 93 output space. 94 */ 95void inflate_fast(strm, start) 96z_streamp strm; 97unsigned start; /* inflate()'s starting value for strm->avail_out */ 98{ 99 struct inflate_state FAR *state; 100 unsigned char FAR *in; /* local strm->next_in */ 101 unsigned char FAR *last; /* while in < last, enough input available */ 102 unsigned char FAR *out; /* local strm->next_out */ 103 unsigned char FAR *beg; /* inflate()'s initial strm->next_out */ 104 unsigned char FAR *end; /* while out < end, enough space available */ 105#ifdef INFLATE_STRICT 106 unsigned dmax; /* maximum distance from zlib header */ 107#endif 108 unsigned wsize; /* window size or zero if not using window */ 109 unsigned whave; /* valid bytes in the window */ 110 unsigned write; /* window write index */ 111 unsigned char FAR *window; /* allocated sliding window, if wsize != 0 */ 112 unsigned long hold; /* local strm->hold */ 113 unsigned bits; /* local strm->bits */ 114 code const FAR *lcode; /* local strm->lencode */ 115 code const FAR *dcode; /* local strm->distcode */ 116 unsigned lmask; /* mask for first level of length codes */ 117 unsigned dmask; /* mask for first level of distance codes */ 118 code this; /* retrieved table entry */ 119 unsigned op; /* code bits, operation, extra bits, or */ 120 /* window position, window bytes to copy */ 121 unsigned len; /* match length, unused bytes */ 122 unsigned dist; /* match distance */ 123 unsigned char FAR *from; /* where to copy match from */ 124 125 /* copy state to local variables */ 126 state = (struct inflate_state FAR *)strm->state; 127 in = strm->next_in - OFF; 128 last = in + (strm->avail_in - 5); 129 out = strm->next_out - OFF; 130 beg = out - (start - strm->avail_out); 131 end = out + (strm->avail_out - 257); 132#ifdef INFLATE_STRICT 133 dmax = state->dmax; 134#endif 135 wsize = state->wsize; 136 whave = state->whave; 137 write = state->write; 138 window = state->window; 139 hold = state->hold; 140 bits = state->bits; 141 lcode = state->lencode; 142 dcode = state->distcode; 143 lmask = (1U << state->lenbits) - 1; 144 dmask = (1U << state->distbits) - 1; 145 146 /* decode literals and length/distances until end-of-block or not enough 147 input data or output space */ 148 do { 149 if (bits < 15) { 150 hold += (unsigned long)(PUP(in)) << bits; 151 bits += 8; 152 hold += (unsigned long)(PUP(in)) << bits; 153 bits += 8; 154 } 155 this = lcode[hold & lmask]; 156 dolen: 157 op = (unsigned)(this.bits); 158 hold >>= op; 159 bits -= op; 160 op = (unsigned)(this.op); 161 if (op == 0) { /* literal */ 162 Tracevv((stderr, this.val >= 0x20 && this.val < 0x7f ? 163 "inflate: literal '%c'\n" : 164 "inflate: literal 0x%02x\n", this.val)); 165 PUP(out) = (unsigned char)(this.val); 166 } 167 else if (op & 16) { /* length base */ 168 len = (unsigned)(this.val); 169 op &= 15; /* number of extra bits */ 170 if (op) { 171 if (bits < op) { 172 hold += (unsigned long)(PUP(in)) << bits; 173 bits += 8; 174 } 175 len += (unsigned)hold & ((1U << op) - 1); 176 hold >>= op; 177 bits -= op; 178 } 179 Tracevv((stderr, "inflate: length %u\n", len)); 180 if (bits < 15) { 181 hold += (unsigned long)(PUP(in)) << bits; 182 bits += 8; 183 hold += (unsigned long)(PUP(in)) << bits; 184 bits += 8; 185 } 186 this = dcode[hold & dmask]; 187 dodist: 188 op = (unsigned)(this.bits); 189 hold >>= op; 190 bits -= op; 191 op = (unsigned)(this.op); 192 if (op & 16) { /* distance base */ 193 dist = (unsigned)(this.val); 194 op &= 15; /* number of extra bits */ 195 if (bits < op) { 196 hold += (unsigned long)(PUP(in)) << bits; 197 bits += 8; 198 if (bits < op) { 199 hold += (unsigned long)(PUP(in)) << bits; 200 bits += 8; 201 } 202 } 203 dist += (unsigned)hold & ((1U << op) - 1); 204#ifdef INFLATE_STRICT 205 if (dist > dmax) { 206 strm->msg = (char *)"invalid distance too far back"; 207 state->mode = BAD; 208 break; 209 } 210#endif 211 hold >>= op; 212 bits -= op; 213 Tracevv((stderr, "inflate: distance %u\n", dist)); 214 op = (unsigned)(out - beg); /* max distance in output */ 215 if (dist > op) { /* see if copy from window */ 216 op = dist - op; /* distance back in window */ 217 if (op > whave) { 218 strm->msg = (char *)"invalid distance too far back"; 219 state->mode = BAD; 220 break; 221 } 222 from = window - OFF; 223 if (write == 0) { /* very common case */ 224 from += wsize - op; 225 if (op < len) { /* some from window */ 226 len -= op; 227 do { 228 PUP(out) = PUP(from); 229 } while (--op); 230 from = out - dist; /* rest from output */ 231 } 232 } 233 else if (write < op) { /* wrap around window */ 234 from += wsize + write - op; 235 op -= write; 236 if (op < len) { /* some from end of window */ 237 len -= op; 238 do { 239 PUP(out) = PUP(from); 240 } while (--op); 241 from = window - OFF; 242 if (write < len) { /* some from start of window */ 243 op = write; 244 len -= op; 245 do { 246 PUP(out) = PUP(from); 247 } while (--op); 248 from = out - dist; /* rest from output */ 249 } 250 } 251 } 252 else { /* contiguous in window */ 253 from += write - op; 254 if (op < len) { /* some from window */ 255 len -= op; 256 do { 257 PUP(out) = PUP(from); 258 } while (--op); 259 from = out - dist; /* rest from output */ 260 } 261 } 262 while (len > 2) { 263 PUP(out) = PUP(from); 264 PUP(out) = PUP(from); 265 PUP(out) = PUP(from); 266 len -= 3; 267 } 268 if (len) { 269 PUP(out) = PUP(from); 270 if (len > 1) 271 PUP(out) = PUP(from); 272 } 273 } 274 else { 275 from = out - dist; /* copy direct from output */ 276 do { /* minimum length is three */ 277 PUP(out) = PUP(from); 278 PUP(out) = PUP(from); 279 PUP(out) = PUP(from); 280 len -= 3; 281 } while (len > 2); 282 if (len) { 283 PUP(out) = PUP(from); 284 if (len > 1) 285 PUP(out) = PUP(from); 286 } 287 } 288 } 289 else if ((op & 64) == 0) { /* 2nd level distance code */ 290 this = dcode[this.val + (hold & ((1U << op) - 1))]; 291 goto dodist; 292 } 293 else { 294 strm->msg = (char *)"invalid distance code"; 295 state->mode = BAD; 296 break; 297 } 298 } 299 else if ((op & 64) == 0) { /* 2nd level length code */ 300 this = lcode[this.val + (hold & ((1U << op) - 1))]; 301 goto dolen; 302 } 303 else if (op & 32) { /* end-of-block */ 304 Tracevv((stderr, "inflate: end of block\n")); 305 state->mode = TYPE; 306 break; 307 } 308 else { 309 strm->msg = (char *)"invalid literal/length code"; 310 state->mode = BAD; 311 break; 312 } 313 } while (in < last && out < end); 314 315 /* return unused bytes (on entry, bits < 8, so in won't go too far back) */ 316 len = bits >> 3; 317 in -= len; 318 bits -= len << 3; 319 hold &= (1U << bits) - 1; 320 321 /* update state and return */ 322 strm->next_in = in + OFF; 323 strm->next_out = out + OFF; 324 strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last)); 325 strm->avail_out = (unsigned)(out < end ? 326 257 + (end - out) : 257 - (out - end)); 327 state->hold = hold; 328 state->bits = bits; 329 return; 330} 331 332/* 333 inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe): 334 - Using bit fields for code structure 335 - Different op definition to avoid & for extra bits (do & for table bits) 336 - Three separate decoding do-loops for direct, window, and write == 0 337 - Special case for distance > 1 copies to do overlapped load and store copy 338 - Explicit branch predictions (based on measured branch probabilities) 339 - Deferring match copy and interspersed it with decoding subsequent codes 340 - Swapping literal/length else 341 - Swapping window/direct else 342 - Larger unrolled copy loops (three is about right) 343 - Moving len -= 3 statement into middle of loop 344 */ 345 346#endif /* !ASMINF */ 347 348