Debian Codegraph @

CHANGES

raw
1
2 libmad - MPEG audio decoder library
3 Copyright (C) 2000-2004 Underbit Technologies, Inc.
4
5 $Id: CHANGES,v 1.14 2004/02/17 02:02:03 rob Exp $
6
7 ===============================================================================
8
9 Verison 0.16.4
10 * Fix compatibility with libmad 0.15.1b
11
12 Version 0.16.3
13 * Disable assembly optimizations for all 64 bit CPU architectures; they are
14 only for 32 and 16 bit architectures
15 * Fix CPU architecture detection for PowerPC
16
17 Version 0.16.2
18 * Fix building assembly file on ARM
19 * Fix pkgconfig file when CMAKE_INSTALL_{INCLUDE,LIB}DIR are absolute paths
20
21 Version 0.16.1
22 * Fix generation of mad.h broken by move to CMake
23 * Add CMake options for CPU architecture-specific optimizations
24
25 Version 0.16.0
26 * Add CMake build system
27 * Remove autotools build system
28 * Add pkgconfig and CMake config files
29 * Apply patches from Debian and Fedora
30
31 Version 0.15.1 (beta)
32
33 * Updated to autoconf 2.59, automake 1.8.2, libtool 1.5.2.
34
35 * Replaced Layer III IMDCT routine with one based on a faster algorithm,
36 improving both speed and accuracy.
37
38 * Improved portability of the Huffman table initialization.
39
40 * Fixed a problem that could result in an assertion failure in layer3.c
41 due to an invalid Layer III free format bitrate.
42
43 * Improved the robustness of Layer II bitrate/mode combinations, and added
44 a new MAD_ERROR_BADMODE error enum. The allowability of low-bitrate
45 stereo streams is influenced by the --enable-strict-iso option to
46 `configure'.
47
48 Version 0.15.0 (beta)
49
50 * Updated to autoconf 2.57, automake 1.7.5, libtool 1.4.3.
51
52 * Added new mad_f_div() API routine.
53
54 * Added a 64th entry to the Layer I/Layer II scalefactor table, for better
55 compatibility with existing streams. The --enable-strict-iso option to
56 `configure' can be used to disable use of this entry.
57
58 * Modified the header decoding routine to allow the reserved emphasis
59 value, for better compatibility with existing streams. The
60 --enable-strict-iso option to `configure' can be used to restore the
61 previous behavior of reporting this value as an error.
62
63 * Added new MAD_EMPHASIS_RESERVED enumeration constant.
64
65 * Fixed a bug in the ARM version of mad_f_scale64() discovered by Andre
66 McCurdy.
67
68 * Rewrote PowerPC assembly for minor gains.
69
70 * Modified mad_timer_fraction() to avoid the possibility of division by
71 zero when 0 is passed as the second argument.
72
73 * Fixed a non-fatal problem caused by attempting to designate ancillary
74 bits in Layer III after a decoding error.
75
76 * Changed to build a shared library by default.
77
78 * Changed to use native Cygwin build by default; give --host=mingw32 to
79 `configure' to use MinGW (and avoid a dependency on the Cygwin DLL).
80
81 Version 0.14.2 (beta)
82
83 * Changed Cygwin builds to use MinGW; resulting Win32 executables no
84 longer have a dependency on Cygwin DLLs.
85
86 * Added a new mad_stream_errorstr() API function to libmad for retrieving
87 a string description of the current error condition.
88
89 Version 0.14.1 (beta)
90
91 * Updated config.guess and config.sub to latest upstream versions.
92
93 * Enabled libtool versioning rather than release numbering.
94
95 * Improved the documentation in minimad.c.
96
97 * Several other small fixes.
98
99 Version 0.14.0 (beta)
100
101 * Added a 64-bit FPM negation operation to improve performance of subband
102 synthesis on some platforms.
103
104 * Improved MSVC++ portability and added MSVC++ project files.
105
106 * Added rounding to Layer III requantization for slightly better accuracy.
107
108 Version 0.13.0 (beta)
109
110 * Ancillary data is now properly extracted from Layer III streams.
111
112 * Rewrote the Layer III joint stereo decoding routine to correct a major
113 MPEG-2 problem and a minor MPEG-1 problem decoding intensity stereo.
114
115 * Eliminated the dependency on sign-extending right shifts for Layer I and
116 Layer II.
117
118 * Renamed `private' field to `private_bits' for better C++ compatibility.
119
120 * Gratuitously renamed `sfreq' field to `samplerate' and
121 MAD_ERROR_BADSAMPLEFREQ constant to MAD_ERROR_BADSAMPLERATE.
122
123 * Added `samplerate' and `channels' fields to synth.pcm struct to allow
124 these to be different from the decoded frame, and for simpler access.
125
126 * Added new mad_stream_options() and mad_decoder_options() API entries for
127 special runtime decoding options.
128
129 * Added new MAD_OPTION_IGNORECRC and MAD_OPTION_HALFSAMPLERATE options.
130
131 * Added new MAD_FLAG_FREEFORMAT indicator flag.
132
133 * Fixed some bugs in the async decoder.
134
135 * Added a new mad_timer_multiply() API routine.
136
137 * Eliminated `+' from asm constraints under Intel for better compatibility
138 with some compilers.
139
140 * Fixed a PIC-related problem in imdct_l_arm.S.
141
142 * Eliminated a static variable to make libmad thread-safe.
143
144 Version 0.12.5 (beta)
145
146 * Modified Layer III requantization to occur during Huffman decoding for
147 significant performance gains.
148
149 * Optimized short block IMDCT by eliminating redundant calculations.
150
151 * Made several other Layer III performance improvements; added
152 ASO_INTERLEAVE1, ASO_INTERLEAVE2, and ASO_ZEROCHECK
153 architecture-specific options for best performance on various
154 architectures.
155
156 * Optimized synthesis DCT to store result values as soon as they are
157 calculated.
158
159 Version 0.12.4 (beta)
160
161 * New PowerPC fixed-point assembly courtesy of David Blythe.
162
163 * Reorganized fixed-point assembly routines for easier maintenance and
164 better performance.
165
166 * Improved performance of subband synthesis through better indexing and
167 fewer local variables.
168
169 * Added alias reduction for the lower two subbands of mixed short blocks,
170 per a report of ambiguity with ISO/IEC 11172-3 and for uniformity with
171 most other implementations. Also improved alias reduction performance
172 using multiply/accumulate.
173
174 * Added --enable-strict-iso option to `configure' to override best
175 accepted practices such as the alias reduction for mixed short blocks.
176
177 * Improved performance of Layer III IMDCT by using longer
178 multiply/accumulate runs where possible.
179
180 Version 0.12.3 (beta)
181
182 * Added MPEG 2.5 support.
183
184 * Added preliminary support for parameterizing the binary point position
185 in the fixed-point representation.
186
187 * Added multiply/accumulate optimization to the Layer III IMDCT for long
188 blocks.
189
190 * Fixed a bug in the handling of Layer III mixed_block_flag.
191
192 * Fixed a configure problem when multiple -O CFLAGS are present.
193
194 Version 0.12.2 (beta)
195
196 * Rearranged the synthesis polyphase filterbank memory vector for better
197 locality of reference, and rewrote mad_synth_frame() to accommodate,
198 resulting in improved performance.
199
200 * Discovered a combination of compiler optimization flags that further
201 improve performance.
202
203 * Changed some array references in layer3.c to pointer derefs.
204
205 Version 0.12.1 (beta)
206
207 * Resolved the intensity + MS joint stereo issue (a simple bug).
208 OPT_ISKLUGE is no longer considered to be a kluge.
209
210 * Fixed another, hopefully last main_data memory bug.
211
212 * Split part of struct mad_frame into struct mad_header for convenience
213 and size.
214
215 Version 0.12.0 (alpha)
216
217 * Changed the build environment to use automake and libtool. A libmad
218 shared library can now be built using the --enable-shared option to
219 `configure'.
220
221 * Added another callback to MAD's high-level decoder API after the frame
222 header has been read but before the frame's audio data is decoded.
223
224 * Streamlined header processing so that mad_frame_decode() can be called
225 with or without having already called mad_frame_header().
226
227 * Fixed some other header reading miscellany, including CRC handling and
228 free bitrate detection, and frame length verification with free
229 bitrates.
230
231 * Fixed a problem with Layer III free bitrates > 320 kbps. The main_data
232 buffer size should now be large enough to handle any size frame, by
233 virtue of the maximum possible part2_3_length.
234
235 * Further developed the async API; arbitrary messages can now be passed to
236 the subsidiary decoding process.
237
238 * Streamlined timer.c and extended its interface. It now has support for
239 video frame/field lengths, including output support for drop-frame
240 encoding.
241
242 * Replaced many constant integer preprocessor defines with enums.
243
244 Version 0.11.4 (beta)
245
246 * Fixed free format bitrate discovery.
247
248 * Changed the timer implementation and extended its interface.
249
250 * Integrated Nicolas Pitre's patch for pre-shifting at compile-time and
251 for better multiply/accumulate code output.
252
253 * Applied Simon Burge's patch to imdct_l_arm.S for a.out compatibility.
254
255 * Added -mtune=strongarm for all ARM targets.
256
257 Version 0.11.3 (beta)
258
259 * Added new --enable-speed and --enable-accuracy options for `configure'
260 to automatically select appropriate SSO/ASO options, et al.
261
262 * Modified subband synthesis to use multiply/accumulate optimization (if
263 available) for better speed and/or accuracy.
264
265 * Incorporated Andre McCurdy's changes for further rounding optimizations
266 in the rest of his code.
267
268 Version 0.11.2 (beta)
269
270 * Incorporated Nicolas Pitre's ARM assembly and parameterized scaling
271 changes.
272
273 * Incorporated Andre McCurdy's ARM assembly optimization (used only if
274 --enable-aso is given to `configure' to enable architecture-specific
275 optimizations.)
276
277 * Reduced FPM_INTEL assembly to two instructions.
278
279 * Fixed accuracy problems with certain FPM modes in synth.c.
280
281 * Improved the accuracy of FPM_APPROX.
282
283 * Improved the accuracy of SSO.
284
285 * Improved sync discovery by checking for a sync word in the following
286 frame.
287
288 * Minor code clean-up.
289
290 * Added experimental rules for generating a libmad.so shared library.
291
292 Version 0.11.1 (beta)
293
294 * Moved libmad code into a separate directory.
295
296 * Changed SSO to be disabled by default, as output accuracy is deemed to
297 be more important than speed in the general case.
298
299 * Fixed a bug in Layer III sanity checking that could cause a crash on
300 certain random data input.
301
302 * Extended the Layer III requantization table from 8191 to 8206 as some
303 encoders are known to use these values, even though ISO/IEC 11172-3
304 suggests the maximum should be 8191.
305
306 Version 0.11.0 (beta)
307
308 * Implemented MPEG-2 extension to Lower Sampling Frequencies.
309
310 * Improved Layer III performance by avoiding IMDCT calculation when all
311 input samples are zero.
312
313 * Significantly reduced size of Layer II tables.
314
315 Version 0.10.3 (beta)
316
317 * Improved SSO output quality.
318
319 * Made portable to cygwin.
320
321 * Localized memory references in III_huffdecode() for better performance.
322
323 Version 0.10.2 (beta)
324
325 * Rewrote Layer III long block 36-point IMDCT routine for better
326 performance.
327
328 * Improved subband synthesis fixed-point games somewhat.
329
330 Version 0.10.1 (beta)
331
332 * Added a subband synthesis optimization (SSO) which involves modifying
333 the fixed-point multiplication method during windowing. This produces
334 subtle differences in the output but improves performance greatly.
335
336 * Added I_STEREO and MS_STEREO flags to frame struct.
337
338 * Eliminated privately-used CRCFAILED flag.
339
340 * Fixed a bug where Layer III decoding could crash on some badly-formatted
341 (e.g. non-MPEG) bitstreams.
342
343 * Miscellaneous code clean-up.
344
345 Version 0.10.0 (beta)
346
347 * Added SPARC fixed-point math support.
348
349 * Revamped libmad API for better high- and low-level support.
350
351 * Documented more of the code.
352
353 * Changed sync semantics such that new stream buffers are assumed to be
354 sync-aligned.
355
356 * Changed Layer III to dynamically allocate static memory so as not to
357 waste it (about 6.4K) when only decoding Layer I or Layer II.
358
359 ===============================================================================
360