|
1 // Copyright (c) 1998-2009 Nokia Corporation and/or its subsidiary(-ies). |
|
2 // All rights reserved. |
|
3 // This component and the accompanying materials are made available |
|
4 // under the terms of "Eclipse Public License v1.0" |
|
5 // which accompanies this distribution, and is available |
|
6 // at the URL "http://www.eclipse.org/legal/epl-v10.html". |
|
7 // |
|
8 // Initial Contributors: |
|
9 // Nokia Corporation - initial contribution. |
|
10 // |
|
11 // Contributors: |
|
12 // |
|
13 // Description: |
|
14 // e32tools\petran\Szip\decode.cpp |
|
15 // |
|
16 // |
|
17 |
|
18 #include "huffman.h" |
|
19 #include "panic.h" |
|
20 #include <cpudefs.h> |
|
21 #include "h_utl.h" |
|
22 #include "farray.h" |
|
23 |
|
24 |
|
25 const TInt KHuffTerminate=0x0001; |
|
26 const TUint32 KBranch1=sizeof(TUint32)<<16; |
|
27 |
|
28 TUint32* HuffmanSubTree(TUint32* aPtr,const TUint32* aValue,TUint32** aLevel) |
|
29 // |
|
30 // write the subtree below aPtr and return the head |
|
31 // |
|
32 { |
|
33 TUint32* l=*aLevel++; |
|
34 if (l>aValue) |
|
35 { |
|
36 TUint32* sub0=HuffmanSubTree(aPtr,aValue,aLevel); // 0-tree first |
|
37 aPtr=HuffmanSubTree(sub0,aValue-(aPtr-sub0)-1,aLevel); // 1-tree |
|
38 TInt branch0=(TUint8*)sub0-(TUint8*)(aPtr-1); |
|
39 *--aPtr=KBranch1|branch0; |
|
40 } |
|
41 else if (l==aValue) |
|
42 { |
|
43 TUint term0=*aValue--; // 0-term |
|
44 aPtr=HuffmanSubTree(aPtr,aValue,aLevel); // 1-tree |
|
45 *--aPtr=KBranch1|(term0>>16); |
|
46 } |
|
47 else // l<iNext |
|
48 { |
|
49 TUint term0=*aValue--; // 0-term |
|
50 TUint term1=*aValue--; |
|
51 *--aPtr=(term1>>16<<16)|(term0>>16); |
|
52 } |
|
53 return aPtr; |
|
54 } |
|
55 |
|
56 void Huffman::Decoding(const TUint32 aHuffman[],TInt aNumCodes,TUint32 aDecodeTree[],TInt aSymbolBase) |
|
57 /** Create a canonical Huffman decoding tree |
|
58 |
|
59 This generates the huffman decoding tree used by TBitInput::HuffmanL() to read huffman |
|
60 encoded data. The input is table of code lengths, as generated by Huffman::HuffmanL() |
|
61 and must represent a valid huffman code. |
|
62 |
|
63 @param "const TUint32 aHuffman[]" The table of code lengths as generated by Huffman::HuffmanL() |
|
64 @param "TInt aNumCodes" The number of codes in the table |
|
65 @param "TUint32 aDecodeTree[]" The space for the decoding tree. This must be the same |
|
66 size as the code-length table, and can safely be the same memory |
|
67 @param "TInt aSymbolBase" the base value for the output 'symbols' from the decoding tree, by default |
|
68 this is zero. |
|
69 |
|
70 @panic "USER ???" If the provided code is not a valid Huffman coding |
|
71 |
|
72 @see IsValid() |
|
73 @see HuffmanL() |
|
74 */ |
|
75 { |
|
76 if(!IsValid(aHuffman,aNumCodes)) |
|
77 Panic(EHuffmanInvalidCoding); |
|
78 // |
|
79 TFixedArray<TInt,KMaxCodeLength> counts; |
|
80 counts.Reset(); |
|
81 TInt codes=0; |
|
82 TInt ii; |
|
83 for (ii=0;ii<aNumCodes;++ii) |
|
84 { |
|
85 TInt len=aHuffman[ii]; |
|
86 aDecodeTree[ii]=len; |
|
87 if (--len>=0) |
|
88 { |
|
89 ++counts[len]; |
|
90 ++codes; |
|
91 } |
|
92 } |
|
93 // |
|
94 TFixedArray<TUint32*,KMaxCodeLength> level; |
|
95 TUint32* lit=aDecodeTree+codes; |
|
96 for (ii=0;ii<KMaxCodeLength;++ii) |
|
97 { |
|
98 level[ii]=lit; |
|
99 lit-=counts[ii]; |
|
100 } |
|
101 aSymbolBase=(aSymbolBase<<17)+(KHuffTerminate<<16); |
|
102 for (ii=0;ii<aNumCodes;++ii) |
|
103 { |
|
104 TUint len=TUint8(aDecodeTree[ii]); |
|
105 if (len) |
|
106 *--level[len-1]|=(ii<<17)+aSymbolBase; |
|
107 } |
|
108 if (codes==1) // codes==1 special case: the tree is not complete |
|
109 { |
|
110 TUint term=aDecodeTree[0]>>16; |
|
111 aDecodeTree[0]=term|(term<<16); // 0- and 1-terminate at root |
|
112 } |
|
113 else if (codes>1) |
|
114 HuffmanSubTree(aDecodeTree+codes-1,aDecodeTree+codes-1,&level[0]); |
|
115 } |
|
116 |
|
117 // The decoding tree for the externalised code |
|
118 const TUint32 HuffmanDecoding[]= |
|
119 { |
|
120 0x0004006c, |
|
121 0x00040064, |
|
122 0x0004005c, |
|
123 0x00040050, |
|
124 0x00040044, |
|
125 0x0004003c, |
|
126 0x00040034, |
|
127 0x00040021, |
|
128 0x00040023, |
|
129 0x00040025, |
|
130 0x00040027, |
|
131 0x00040029, |
|
132 0x00040014, |
|
133 0x0004000c, |
|
134 0x00040035, |
|
135 0x00390037, |
|
136 0x00330031, |
|
137 0x0004002b, |
|
138 0x002f002d, |
|
139 0x001f001d, |
|
140 0x001b0019, |
|
141 0x00040013, |
|
142 0x00170015, |
|
143 0x0004000d, |
|
144 0x0011000f, |
|
145 0x000b0009, |
|
146 0x00070003, |
|
147 0x00050001 |
|
148 }; |
|
149 |
|
150 void Huffman::InternalizeL(TBitInput& aInput,TUint32 aHuffman[],TInt aNumCodes) |
|
151 /** Restore a canonical huffman encoding from a bit stream |
|
152 |
|
153 The encoding must have been stored using Huffman::ExternalizeL(). The resulting |
|
154 code-length table can be used to create an encoding table using Huffman::Encoding() |
|
155 or a decoding tree using Huffman::Decoding(). |
|
156 |
|
157 @param "TBitInput& aInput" The input stream with the encoding |
|
158 @param "TUint32 aHuffman[]" The internalized code-length table is placed here |
|
159 @param "TInt aNumCodes" The number of huffman codes in the table |
|
160 |
|
161 @leave "TBitInput::HuffmanL()" |
|
162 |
|
163 @see ExternalizeL() |
|
164 */ |
|
165 // See ExternalizeL for a description of the format |
|
166 { |
|
167 // initialise move-to-front list |
|
168 TFixedArray<TUint8,Huffman::KMetaCodes> list; |
|
169 for (TInt i=0;i<list.Count();++i) |
|
170 list[i]=TUint8(i); |
|
171 TInt last=0; |
|
172 // extract codes, reverse rle-0 and mtf encoding in one pass |
|
173 TUint32* p=aHuffman; |
|
174 const TUint32* end=aHuffman+aNumCodes; |
|
175 TInt rl=0; |
|
176 while (p+rl<end) |
|
177 { |
|
178 TInt c=aInput.HuffmanL(HuffmanDecoding); |
|
179 if (c<2) |
|
180 { |
|
181 // one of the zero codes used by RLE-0 |
|
182 // update he run-length |
|
183 rl+=rl+c+1; |
|
184 } |
|
185 else |
|
186 { |
|
187 while (rl>0) |
|
188 { |
|
189 if (p>end) |
|
190 { |
|
191 Panic(EHuffmanCorruptFile); |
|
192 } |
|
193 *p++=last; |
|
194 --rl; |
|
195 } |
|
196 --c; |
|
197 list[0]=TUint8(last); |
|
198 last=list[c]; |
|
199 HMem::Copy(&list[1],&list[0],c); |
|
200 if (p>end) |
|
201 { |
|
202 Panic(EHuffmanCorruptFile); |
|
203 } |
|
204 *p++=last; |
|
205 } |
|
206 } |
|
207 while (rl>0) |
|
208 { |
|
209 if (p>end) |
|
210 { |
|
211 Panic(EHuffmanCorruptFile); |
|
212 } |
|
213 *p++=last; |
|
214 --rl; |
|
215 } |
|
216 } |
|
217 |
|
218 // bit-stream input class |
|
219 |
|
220 inline TUint reverse(TUint aVal) |
|
221 // |
|
222 // Reverse the byte-order of a 32 bit value |
|
223 // This generates optimal ARM code (4 instructions) |
|
224 // |
|
225 { |
|
226 TUint v=(aVal<<16)|(aVal>>16); |
|
227 v^=aVal; |
|
228 v&=0xff00ffff; |
|
229 aVal=(aVal>>8)|(aVal<<24); |
|
230 return aVal^(v>>8); |
|
231 } |
|
232 |
|
233 TBitInput::TBitInput() |
|
234 /** Construct a bit stream input object |
|
235 |
|
236 Following construction the bit stream is ready for reading bits, but will |
|
237 immediately call UnderflowL() as the input buffer is empty. |
|
238 */ |
|
239 :iCount(0),iRemain(0) |
|
240 {} |
|
241 |
|
242 TBitInput::TBitInput(const TUint8* aPtr, TInt aLength, TInt aOffset) |
|
243 /** Construct a bit stream input object over a buffer |
|
244 |
|
245 Following construction the bit stream is ready for reading bits from |
|
246 the specified buffer. |
|
247 |
|
248 @param "const TUint8* aPtr" The address of the buffer containing the bit stream |
|
249 @param "TInt aLength" The length of the bitstream in bits |
|
250 @param "TInt aOffset" The bit offset from the start of the buffer to the bit stream (defaults to zero) |
|
251 */ |
|
252 { |
|
253 Set(aPtr,aLength,aOffset); |
|
254 } |
|
255 |
|
256 void TBitInput::Set(const TUint8* aPtr, TInt aLength, TInt aOffset) |
|
257 /** Set the memory buffer to use for input |
|
258 |
|
259 Bits will be read from this buffer until it is empty, at which point |
|
260 UnderflowL() will be called. |
|
261 |
|
262 @param "const TUint8* aPtr" The address of the buffer containing the bit stream |
|
263 @param "TInt aLength" The length of the bitstream in bits |
|
264 @param "TInt aOffset" The bit offset from the start of the buffer to the bit stream (defaults to zero) |
|
265 */ |
|
266 { |
|
267 TUint p=(TUint)aPtr; |
|
268 p+=aOffset>>3; // nearest byte to the specified bit offset |
|
269 aOffset&=7; // bit offset within the byte |
|
270 const TUint32* ptr=(const TUint32*)(p&~3); // word containing this byte |
|
271 aOffset+=(p&3)<<3; // bit offset within the word |
|
272 if (aLength==0) |
|
273 iCount=0; |
|
274 else |
|
275 { |
|
276 // read the first few bits of the stream |
|
277 iBits=reverse(*ptr++)<<aOffset; |
|
278 aOffset=32-aOffset; |
|
279 aLength-=aOffset; |
|
280 if (aLength<0) |
|
281 aOffset+=aLength; |
|
282 iCount=aOffset; |
|
283 } |
|
284 iRemain=aLength; |
|
285 iPtr=ptr; |
|
286 } |
|
287 |
|
288 #ifndef __HUFFMAN_MACHINE_CODED__ |
|
289 |
|
290 TUint TBitInput::ReadL() |
|
291 /** Read a single bit from the input |
|
292 |
|
293 Return the next bit in the input stream. This will call UnderflowL() if |
|
294 there are no more bits available. |
|
295 |
|
296 @return The next bit in the stream |
|
297 |
|
298 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
299 to get more data |
|
300 */ |
|
301 { |
|
302 TInt c=iCount; |
|
303 TUint bits=iBits; |
|
304 if (--c<0) |
|
305 return ReadL(1); |
|
306 iCount=c; |
|
307 iBits=bits<<1; |
|
308 return bits>>31; |
|
309 } |
|
310 |
|
311 TUint TBitInput::ReadL(TInt aSize) |
|
312 /** Read a multi-bit value from the input |
|
313 |
|
314 Return the next few bits as an unsigned integer. The last bit read is |
|
315 the least significant bit of the returned value, and the value is |
|
316 zero extended to return a 32-bit result. |
|
317 |
|
318 A read of zero bits will always reaturn zero. |
|
319 |
|
320 This will call UnderflowL() if there are not enough bits available. |
|
321 |
|
322 @param "TInt aSize" The number of bits to read |
|
323 |
|
324 @return The bits read from the stream |
|
325 |
|
326 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
327 to get more data |
|
328 */ |
|
329 { |
|
330 if (!aSize) |
|
331 return 0; |
|
332 TUint val=0; |
|
333 TUint bits=iBits; |
|
334 iCount-=aSize; |
|
335 while (iCount<0) |
|
336 { |
|
337 // need more bits |
|
338 #ifdef __CPU_X86 |
|
339 // X86 does not allow shift-by-32 |
|
340 if (iCount+aSize!=0) |
|
341 val|=bits>>(32-(iCount+aSize))<<(-iCount); // scrub low order bits |
|
342 #else |
|
343 val|=bits>>(32-(iCount+aSize))<<(-iCount); // scrub low order bits |
|
344 #endif |
|
345 aSize=-iCount; // bits still required |
|
346 if (iRemain>0) |
|
347 { |
|
348 bits=reverse(*iPtr++); |
|
349 iCount+=32; |
|
350 iRemain-=32; |
|
351 if (iRemain<0) |
|
352 iCount+=iRemain; |
|
353 } |
|
354 else |
|
355 { |
|
356 UnderflowL(); |
|
357 bits=iBits; |
|
358 iCount-=aSize; |
|
359 } |
|
360 } |
|
361 #ifdef __CPU_X86 |
|
362 // X86 does not allow shift-by-32 |
|
363 iBits=aSize==32?0:bits<<aSize; |
|
364 #else |
|
365 iBits=bits<<aSize; |
|
366 #endif |
|
367 return val|(bits>>(32-aSize)); |
|
368 } |
|
369 |
|
370 TUint TBitInput::HuffmanL(const TUint32* aTree) |
|
371 /** Read and decode a Huffman Code |
|
372 |
|
373 Interpret the next bits in the input as a Huffman code in the specified |
|
374 decoding. The decoding tree should be the output from Huffman::Decoding(). |
|
375 |
|
376 @param "const TUint32* aTree" The huffman decoding tree |
|
377 |
|
378 @return The symbol that was decoded |
|
379 |
|
380 @leave "UnderflowL()" It the bit stream is exhausted more UnderflowL is called |
|
381 to get more data |
|
382 */ |
|
383 { |
|
384 TUint huff=0; |
|
385 do |
|
386 { |
|
387 aTree=PtrAdd(aTree,huff>>16); |
|
388 huff=*aTree; |
|
389 if (ReadL()==0) |
|
390 huff<<=16; |
|
391 } while ((huff&0x10000u)==0); |
|
392 return huff>>17; |
|
393 } |
|
394 |
|
395 #endif |
|
396 |
|
397 void TBitInput::UnderflowL() |
|
398 /** Handle an empty input buffer |
|
399 |
|
400 This virtual function is called when the input buffer is empty and |
|
401 more bits are required. It should reset the input buffer with more |
|
402 data using Set(). |
|
403 |
|
404 A derived class can replace this to read the data from a file |
|
405 (for example) before reseting the input buffer. |
|
406 |
|
407 @leave "KErrUnderflow" The default implementation leaves |
|
408 */ |
|
409 { |
|
410 Panic(EHuffmanBufferOverflow); |
|
411 } |
|
412 |