| OLD | NEW |
| 1 /* inffast.c -- fast decoding | 1 /* inffast.c -- fast decoding |
| 2 * Copyright (C) 1995-2004 Mark Adler | 2 * Copyright (C) 1995-2008, 2010 Mark Adler |
| 3 * For conditions of distribution and use, see copyright notice in zlib.h | 3 * For conditions of distribution and use, see copyright notice in zlib.h |
| 4 */ | 4 */ |
| 5 | 5 |
| 6 #include "zutil.h" | 6 #include "zutil.h" |
| 7 #include "inftrees.h" | 7 #include "inftrees.h" |
| 8 #include "inflate.h" | 8 #include "inflate.h" |
| 9 #include "inffast.h" | 9 #include "inffast.h" |
| 10 | 10 |
| 11 #ifndef ASMINF | 11 #ifndef ASMINF |
| 12 | 12 |
| (...skipping 44 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
| 57 length code, 5 bits for the length extra, 15 bits for the distance code, | 57 length code, 5 bits for the length extra, 15 bits for the distance code, |
| 58 and 13 bits for the distance extra. This totals 48 bits, or six bytes. | 58 and 13 bits for the distance extra. This totals 48 bits, or six bytes. |
| 59 Therefore if strm->avail_in >= 6, then there is enough input to avoid | 59 Therefore if strm->avail_in >= 6, then there is enough input to avoid |
| 60 checking for available input while decoding. | 60 checking for available input while decoding. |
| 61 | 61 |
| 62 - The maximum bytes that a single length/distance pair can output is 258 | 62 - The maximum bytes that a single length/distance pair can output is 258 |
| 63 bytes, which is the maximum length that can be coded. inflate_fast() | 63 bytes, which is the maximum length that can be coded. inflate_fast() |
| 64 requires strm->avail_out >= 258 for each loop to avoid checking for | 64 requires strm->avail_out >= 258 for each loop to avoid checking for |
| 65 output space. | 65 output space. |
| 66 */ | 66 */ |
| 67 void inflate_fast(strm, start) | 67 void ZLIB_INTERNAL inflate_fast(strm, start) |
| 68 z_streamp strm; | 68 z_streamp strm; |
| 69 unsigned start; /* inflate()'s starting value for strm->avail_out */ | 69 unsigned start; /* inflate()'s starting value for strm->avail_out */ |
| 70 { | 70 { |
| 71 struct inflate_state FAR *state; | 71 struct inflate_state FAR *state; |
| 72 unsigned char FAR *in; /* local strm->next_in */ | 72 unsigned char FAR *in; /* local strm->next_in */ |
| 73 unsigned char FAR *last; /* while in < last, enough input available */ | 73 unsigned char FAR *last; /* while in < last, enough input available */ |
| 74 unsigned char FAR *out; /* local strm->next_out */ | 74 unsigned char FAR *out; /* local strm->next_out */ |
| 75 unsigned char FAR *beg; /* inflate()'s initial strm->next_out */ | 75 unsigned char FAR *beg; /* inflate()'s initial strm->next_out */ |
| 76 unsigned char FAR *end; /* while out < end, enough space available */ | 76 unsigned char FAR *end; /* while out < end, enough space available */ |
| 77 #ifdef INFLATE_STRICT | 77 #ifdef INFLATE_STRICT |
| 78 unsigned dmax; /* maximum distance from zlib header */ | 78 unsigned dmax; /* maximum distance from zlib header */ |
| 79 #endif | 79 #endif |
| 80 unsigned wsize; /* window size or zero if not using window */ | 80 unsigned wsize; /* window size or zero if not using window */ |
| 81 unsigned whave; /* valid bytes in the window */ | 81 unsigned whave; /* valid bytes in the window */ |
| 82 unsigned write; /* window write index */ | 82 unsigned wnext; /* window write index */ |
| 83 unsigned char FAR *window; /* allocated sliding window, if wsize != 0 */ | 83 unsigned char FAR *window; /* allocated sliding window, if wsize != 0 */ |
| 84 unsigned long hold; /* local strm->hold */ | 84 unsigned long hold; /* local strm->hold */ |
| 85 unsigned bits; /* local strm->bits */ | 85 unsigned bits; /* local strm->bits */ |
| 86 code const FAR *lcode; /* local strm->lencode */ | 86 code const FAR *lcode; /* local strm->lencode */ |
| 87 code const FAR *dcode; /* local strm->distcode */ | 87 code const FAR *dcode; /* local strm->distcode */ |
| 88 unsigned lmask; /* mask for first level of length codes */ | 88 unsigned lmask; /* mask for first level of length codes */ |
| 89 unsigned dmask; /* mask for first level of distance codes */ | 89 unsigned dmask; /* mask for first level of distance codes */ |
| 90 code this; /* retrieved table entry */ | 90 code here; /* retrieved table entry */ |
| 91 unsigned op; /* code bits, operation, extra bits, or */ | 91 unsigned op; /* code bits, operation, extra bits, or */ |
| 92 /* window position, window bytes to copy */ | 92 /* window position, window bytes to copy */ |
| 93 unsigned len; /* match length, unused bytes */ | 93 unsigned len; /* match length, unused bytes */ |
| 94 unsigned dist; /* match distance */ | 94 unsigned dist; /* match distance */ |
| 95 unsigned char FAR *from; /* where to copy match from */ | 95 unsigned char FAR *from; /* where to copy match from */ |
| 96 | 96 |
| 97 /* copy state to local variables */ | 97 /* copy state to local variables */ |
| 98 state = (struct inflate_state FAR *)strm->state; | 98 state = (struct inflate_state FAR *)strm->state; |
| 99 in = strm->next_in - OFF; | 99 in = strm->next_in - OFF; |
| 100 last = in + (strm->avail_in - 5); | 100 last = in + (strm->avail_in - 5); |
| 101 out = strm->next_out - OFF; | 101 out = strm->next_out - OFF; |
| 102 beg = out - (start - strm->avail_out); | 102 beg = out - (start - strm->avail_out); |
| 103 end = out + (strm->avail_out - 257); | 103 end = out + (strm->avail_out - 257); |
| 104 #ifdef INFLATE_STRICT | 104 #ifdef INFLATE_STRICT |
| 105 dmax = state->dmax; | 105 dmax = state->dmax; |
| 106 #endif | 106 #endif |
| 107 wsize = state->wsize; | 107 wsize = state->wsize; |
| 108 whave = state->whave; | 108 whave = state->whave; |
| 109 write = state->write; | 109 wnext = state->wnext; |
| 110 window = state->window; | 110 window = state->window; |
| 111 hold = state->hold; | 111 hold = state->hold; |
| 112 bits = state->bits; | 112 bits = state->bits; |
| 113 lcode = state->lencode; | 113 lcode = state->lencode; |
| 114 dcode = state->distcode; | 114 dcode = state->distcode; |
| 115 lmask = (1U << state->lenbits) - 1; | 115 lmask = (1U << state->lenbits) - 1; |
| 116 dmask = (1U << state->distbits) - 1; | 116 dmask = (1U << state->distbits) - 1; |
| 117 | 117 |
| 118 /* decode literals and length/distances until end-of-block or not enough | 118 /* decode literals and length/distances until end-of-block or not enough |
| 119 input data or output space */ | 119 input data or output space */ |
| 120 do { | 120 do { |
| 121 if (bits < 15) { | 121 if (bits < 15) { |
| 122 hold += (unsigned long)(PUP(in)) << bits; | 122 hold += (unsigned long)(PUP(in)) << bits; |
| 123 bits += 8; | 123 bits += 8; |
| 124 hold += (unsigned long)(PUP(in)) << bits; | 124 hold += (unsigned long)(PUP(in)) << bits; |
| 125 bits += 8; | 125 bits += 8; |
| 126 } | 126 } |
| 127 this = lcode[hold & lmask]; | 127 here = lcode[hold & lmask]; |
| 128 dolen: | 128 dolen: |
| 129 op = (unsigned)(this.bits); | 129 op = (unsigned)(here.bits); |
| 130 hold >>= op; | 130 hold >>= op; |
| 131 bits -= op; | 131 bits -= op; |
| 132 op = (unsigned)(this.op); | 132 op = (unsigned)(here.op); |
| 133 if (op == 0) { /* literal */ | 133 if (op == 0) { /* literal */ |
| 134 Tracevv((stderr, this.val >= 0x20 && this.val < 0x7f ? | 134 Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ? |
| 135 "inflate: literal '%c'\n" : | 135 "inflate: literal '%c'\n" : |
| 136 "inflate: literal 0x%02x\n", this.val)); | 136 "inflate: literal 0x%02x\n", here.val)); |
| 137 PUP(out) = (unsigned char)(this.val); | 137 PUP(out) = (unsigned char)(here.val); |
| 138 } | 138 } |
| 139 else if (op & 16) { /* length base */ | 139 else if (op & 16) { /* length base */ |
| 140 len = (unsigned)(this.val); | 140 len = (unsigned)(here.val); |
| 141 op &= 15; /* number of extra bits */ | 141 op &= 15; /* number of extra bits */ |
| 142 if (op) { | 142 if (op) { |
| 143 if (bits < op) { | 143 if (bits < op) { |
| 144 hold += (unsigned long)(PUP(in)) << bits; | 144 hold += (unsigned long)(PUP(in)) << bits; |
| 145 bits += 8; | 145 bits += 8; |
| 146 } | 146 } |
| 147 len += (unsigned)hold & ((1U << op) - 1); | 147 len += (unsigned)hold & ((1U << op) - 1); |
| 148 hold >>= op; | 148 hold >>= op; |
| 149 bits -= op; | 149 bits -= op; |
| 150 } | 150 } |
| 151 Tracevv((stderr, "inflate: length %u\n", len)); | 151 Tracevv((stderr, "inflate: length %u\n", len)); |
| 152 if (bits < 15) { | 152 if (bits < 15) { |
| 153 hold += (unsigned long)(PUP(in)) << bits; | 153 hold += (unsigned long)(PUP(in)) << bits; |
| 154 bits += 8; | 154 bits += 8; |
| 155 hold += (unsigned long)(PUP(in)) << bits; | 155 hold += (unsigned long)(PUP(in)) << bits; |
| 156 bits += 8; | 156 bits += 8; |
| 157 } | 157 } |
| 158 this = dcode[hold & dmask]; | 158 here = dcode[hold & dmask]; |
| 159 dodist: | 159 dodist: |
| 160 op = (unsigned)(this.bits); | 160 op = (unsigned)(here.bits); |
| 161 hold >>= op; | 161 hold >>= op; |
| 162 bits -= op; | 162 bits -= op; |
| 163 op = (unsigned)(this.op); | 163 op = (unsigned)(here.op); |
| 164 if (op & 16) { /* distance base */ | 164 if (op & 16) { /* distance base */ |
| 165 dist = (unsigned)(this.val); | 165 dist = (unsigned)(here.val); |
| 166 op &= 15; /* number of extra bits */ | 166 op &= 15; /* number of extra bits */ |
| 167 if (bits < op) { | 167 if (bits < op) { |
| 168 hold += (unsigned long)(PUP(in)) << bits; | 168 hold += (unsigned long)(PUP(in)) << bits; |
| 169 bits += 8; | 169 bits += 8; |
| 170 if (bits < op) { | 170 if (bits < op) { |
| 171 hold += (unsigned long)(PUP(in)) << bits; | 171 hold += (unsigned long)(PUP(in)) << bits; |
| 172 bits += 8; | 172 bits += 8; |
| 173 } | 173 } |
| 174 } | 174 } |
| 175 dist += (unsigned)hold & ((1U << op) - 1); | 175 dist += (unsigned)hold & ((1U << op) - 1); |
| 176 #ifdef INFLATE_STRICT | 176 #ifdef INFLATE_STRICT |
| 177 if (dist > dmax) { | 177 if (dist > dmax) { |
| 178 strm->msg = (char *)"invalid distance too far back"; | 178 strm->msg = (char *)"invalid distance too far back"; |
| 179 state->mode = BAD; | 179 state->mode = BAD; |
| 180 break; | 180 break; |
| 181 } | 181 } |
| 182 #endif | 182 #endif |
| 183 hold >>= op; | 183 hold >>= op; |
| 184 bits -= op; | 184 bits -= op; |
| 185 Tracevv((stderr, "inflate: distance %u\n", dist)); | 185 Tracevv((stderr, "inflate: distance %u\n", dist)); |
| 186 op = (unsigned)(out - beg); /* max distance in output */ | 186 op = (unsigned)(out - beg); /* max distance in output */ |
| 187 if (dist > op) { /* see if copy from window */ | 187 if (dist > op) { /* see if copy from window */ |
| 188 op = dist - op; /* distance back in window */ | 188 op = dist - op; /* distance back in window */ |
| 189 if (op > whave) { | 189 if (op > whave) { |
| 190 strm->msg = (char *)"invalid distance too far back"; | 190 if (state->sane) { |
| 191 state->mode = BAD; | 191 strm->msg = |
| 192 break; | 192 (char *)"invalid distance too far back"; |
| 193 state->mode = BAD; |
| 194 break; |
| 195 } |
| 196 #ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR |
| 197 if (len <= op - whave) { |
| 198 do { |
| 199 PUP(out) = 0; |
| 200 } while (--len); |
| 201 continue; |
| 202 } |
| 203 len -= op - whave; |
| 204 do { |
| 205 PUP(out) = 0; |
| 206 } while (--op > whave); |
| 207 if (op == 0) { |
| 208 from = out - dist; |
| 209 do { |
| 210 PUP(out) = PUP(from); |
| 211 } while (--len); |
| 212 continue; |
| 213 } |
| 214 #endif |
| 193 } | 215 } |
| 194 from = window - OFF; | 216 from = window - OFF; |
| 195 if (write == 0) { /* very common case */ | 217 if (wnext == 0) { /* very common case */ |
| 196 from += wsize - op; | 218 from += wsize - op; |
| 197 if (op < len) { /* some from window */ | 219 if (op < len) { /* some from window */ |
| 198 len -= op; | 220 len -= op; |
| 199 do { | 221 do { |
| 200 PUP(out) = PUP(from); | 222 PUP(out) = PUP(from); |
| 201 } while (--op); | 223 } while (--op); |
| 202 from = out - dist; /* rest from output */ | 224 from = out - dist; /* rest from output */ |
| 203 } | 225 } |
| 204 } | 226 } |
| 205 else if (write < op) { /* wrap around window */ | 227 else if (wnext < op) { /* wrap around window */ |
| 206 from += wsize + write - op; | 228 from += wsize + wnext - op; |
| 207 op -= write; | 229 op -= wnext; |
| 208 if (op < len) { /* some from end of window */ | 230 if (op < len) { /* some from end of window */ |
| 209 len -= op; | 231 len -= op; |
| 210 do { | 232 do { |
| 211 PUP(out) = PUP(from); | 233 PUP(out) = PUP(from); |
| 212 } while (--op); | 234 } while (--op); |
| 213 from = window - OFF; | 235 from = window - OFF; |
| 214 if (write < len) { /* some from start of window */ | 236 if (wnext < len) { /* some from start of window */ |
| 215 op = write; | 237 op = wnext; |
| 216 len -= op; | 238 len -= op; |
| 217 do { | 239 do { |
| 218 PUP(out) = PUP(from); | 240 PUP(out) = PUP(from); |
| 219 } while (--op); | 241 } while (--op); |
| 220 from = out - dist; /* rest from output */ | 242 from = out - dist; /* rest from output */ |
| 221 } | 243 } |
| 222 } | 244 } |
| 223 } | 245 } |
| 224 else { /* contiguous in window */ | 246 else { /* contiguous in window */ |
| 225 from += write - op; | 247 from += wnext - op; |
| 226 if (op < len) { /* some from window */ | 248 if (op < len) { /* some from window */ |
| 227 len -= op; | 249 len -= op; |
| 228 do { | 250 do { |
| 229 PUP(out) = PUP(from); | 251 PUP(out) = PUP(from); |
| 230 } while (--op); | 252 } while (--op); |
| 231 from = out - dist; /* rest from output */ | 253 from = out - dist; /* rest from output */ |
| 232 } | 254 } |
| 233 } | 255 } |
| 234 while (len > 2) { | 256 while (len > 2) { |
| 235 PUP(out) = PUP(from); | 257 PUP(out) = PUP(from); |
| (...skipping 16 matching lines...) Expand all Loading... |
| 252 len -= 3; | 274 len -= 3; |
| 253 } while (len > 2); | 275 } while (len > 2); |
| 254 if (len) { | 276 if (len) { |
| 255 PUP(out) = PUP(from); | 277 PUP(out) = PUP(from); |
| 256 if (len > 1) | 278 if (len > 1) |
| 257 PUP(out) = PUP(from); | 279 PUP(out) = PUP(from); |
| 258 } | 280 } |
| 259 } | 281 } |
| 260 } | 282 } |
| 261 else if ((op & 64) == 0) { /* 2nd level distance code */ | 283 else if ((op & 64) == 0) { /* 2nd level distance code */ |
| 262 this = dcode[this.val + (hold & ((1U << op) - 1))]; | 284 here = dcode[here.val + (hold & ((1U << op) - 1))]; |
| 263 goto dodist; | 285 goto dodist; |
| 264 } | 286 } |
| 265 else { | 287 else { |
| 266 strm->msg = (char *)"invalid distance code"; | 288 strm->msg = (char *)"invalid distance code"; |
| 267 state->mode = BAD; | 289 state->mode = BAD; |
| 268 break; | 290 break; |
| 269 } | 291 } |
| 270 } | 292 } |
| 271 else if ((op & 64) == 0) { /* 2nd level length code */ | 293 else if ((op & 64) == 0) { /* 2nd level length code */ |
| 272 this = lcode[this.val + (hold & ((1U << op) - 1))]; | 294 here = lcode[here.val + (hold & ((1U << op) - 1))]; |
| 273 goto dolen; | 295 goto dolen; |
| 274 } | 296 } |
| 275 else if (op & 32) { /* end-of-block */ | 297 else if (op & 32) { /* end-of-block */ |
| 276 Tracevv((stderr, "inflate: end of block\n")); | 298 Tracevv((stderr, "inflate: end of block\n")); |
| 277 state->mode = TYPE; | 299 state->mode = TYPE; |
| 278 break; | 300 break; |
| 279 } | 301 } |
| 280 else { | 302 else { |
| 281 strm->msg = (char *)"invalid literal/length code"; | 303 strm->msg = (char *)"invalid literal/length code"; |
| 282 state->mode = BAD; | 304 state->mode = BAD; |
| (...skipping 15 matching lines...) Expand all Loading... |
| 298 257 + (end - out) : 257 - (out - end)); | 320 257 + (end - out) : 257 - (out - end)); |
| 299 state->hold = hold; | 321 state->hold = hold; |
| 300 state->bits = bits; | 322 state->bits = bits; |
| 301 return; | 323 return; |
| 302 } | 324 } |
| 303 | 325 |
| 304 /* | 326 /* |
| 305 inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe): | 327 inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe): |
| 306 - Using bit fields for code structure | 328 - Using bit fields for code structure |
| 307 - Different op definition to avoid & for extra bits (do & for table bits) | 329 - Different op definition to avoid & for extra bits (do & for table bits) |
| 308 - Three separate decoding do-loops for direct, window, and write == 0 | 330 - Three separate decoding do-loops for direct, window, and wnext == 0 |
| 309 - Special case for distance > 1 copies to do overlapped load and store copy | 331 - Special case for distance > 1 copies to do overlapped load and store copy |
| 310 - Explicit branch predictions (based on measured branch probabilities) | 332 - Explicit branch predictions (based on measured branch probabilities) |
| 311 - Deferring match copy and interspersed it with decoding subsequent codes | 333 - Deferring match copy and interspersed it with decoding subsequent codes |
| 312 - Swapping literal/length else | 334 - Swapping literal/length else |
| 313 - Swapping window/direct else | 335 - Swapping window/direct else |
| 314 - Larger unrolled copy loops (three is about right) | 336 - Larger unrolled copy loops (three is about right) |
| 315 - Moving len -= 3 statement into middle of loop | 337 - Moving len -= 3 statement into middle of loop |
| 316 */ | 338 */ |
| 317 | 339 |
| 318 #endif /* !ASMINF */ | 340 #endif /* !ASMINF */ |
| OLD | NEW |