Debian Codegraph @

doc/html/history.html

raw
1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
2 "http://www.w3.org/TR/html4/loose.dtd">
3 <html>
4 <head>
5 <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
6 <meta name="GENERATOR" content="Mozilla/4.7 [en] (X11; I; Linux 2.2.12-20 i686) [Netscape]">
7 <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
8 <title>LAME Changelog</title>
9 </head>
10 <body style="color: rgb(0, 0, 0); background-color: rgb(255, 255, 255);" alink="#bbbbbb" link="#888888" vlink="#555555">
11 <center>
12 <h1>History</h1>
13 </center>
14 Starting with LAME 3.0: <br>
15 <font color="#ff0000">red = features and bug fixes which affect quality</font> <br>
16 <font color="#3366ff">blue = features and bug fixes which affect speed</font> <br>
17 black = usability, portability, other
18 <hr>
19 <h3>LAME 3.101 &nbsp;under construction</h3>
20 <ul>
21 <li>
22 Robert Hegemann
23 <ul>
24 <li>Patch submitted by KO Myung-Hun, patch ticket <i>[ #80 ] OS/2 patches</i>
25 </li>
26 <li>Patch submitted by Elio Blanca, patch ticket <i>[ #82 ] Take advantage of terminal width on printing file names</i>
27 </li>
28 <li><font color="#ff0000">
29 Bug fix for item <i>[ #496 ] A critical bug in init_xrpow_core_sse</i>
30 </font>
31 </li>
32 <li><font color="#ff0000">
33 Bug fix for item <i>[ #500 ] Buffer overflow in encoder</i>
34 </font>
35 </li>
36 <li><font color="#ff0000">
37 Bug fix for item <i>[ #501 ] Encoder: Assertion 'eov->bitrate_index &lt;&#61; cfg->vbr_max_bitrate_index' failed during ABR encoding</i>
38 </font>
39 </li>
40 <li>Bug fix for item <i>[ #444 ] msacmdrv.h (structure packing alignment)</i>
41 </li>
42 </ul>
43 </li>
44 <li>
45 Alexander Leidinger
46 <ul>
47 <li><font color="#3366ff">Add faster CRC routine. The speed improvement is very small for typical use cases, but may be valuable if a lot of independent encodes are running and/or for a very long time. Patch by Robert Kausch of the fre:ac project.
48 </font>
49 </li>
50 <li><font color="#3366ff">Fix configure script glob-ranges matching of compiler versions. This may result in faster code if no compiler optimization flags are specified during the configure step. This is part of patch ticket <i>[ #491 ] lame 3.100 slower than 3.99.5</i>
51 </font>
52 </li>
53 <li><font color="#3366ff">Disable Takehiros IEEE753 hack by default. On modern CPUs (anything more recent from AMD than hammer/k8; Intel: Core2, i3/5/7/9 and similar -- no idea about recent Atom/Pentium) it is a speed pessimization. Add a configure option for it so that it can be enabled on old CPUs.
54 </font>
55 </li>
56 <li>Update to more recent autotools based scripts, this may or may not fix issues during the configure stage for less popular or more recent OS or architectures.
57 </li>
58 <li>Use external libmpg123 instead of internal mpglib for mpeg decoding (unix-like systems which use the autotools ("configure; make; make install") build system). There are years of improvements in libmpg123 which we do not have in mpglib. Patch by Thomas Orgis of the mpg123 project.
59 </li>
60 <li>Update the Visual Studio project files to Visual Studio 2019. Patch by Michel Fink.
61 </li>
62 <li>Remove the macosx XCode project files, nobody stepped up to update them for the libmpg123 changes.
63 </li>
64 <li>Add ID3v2.4 support (UTF-8 ID3 tags). Patch by "kris".
65 </li>
66 <li>Add pkg-config support. Patch by Nicolas Boulenguez(Debian).
67 </li>
68 <li>Add IPv6 support for mp3rtp. This includes an incompatible change to the command line arguments for existing use cases / scripts. Patch by Surabhi.
69 </li>
70 </ul>
71 </li>
72 </ul>
73 <br>
74 <h3>LAME 3.100 &nbsp;October 13 2017</h3>
75 <ul>
76 <li>
77 Rog&eacute;rio Brito
78 <ul>
79 <li>
80 Don't include the debian directory as one that is needed during
81 builds. Patch taken from Debian's packaging of lame.
82 </li>
83 <li>
84 Resurrect Owen Taylor's code dated from 97-11-3 to properly deal
85 with GTK1. This was transplanted back from aclocal.m4 with a patch
86 provided by Andres Mejia. This change makes it easy to regenerate
87 autotools' files with a simple invocation of <code>autoconf -vfi</code>.
88 </li>
89 <li>
90 Fix possible race condition causing build failures in libmp3lame.
91 Discovered in automated builds by the Debian project with patch
92 provided by Andres Mejia.
93 </li>
94 </ul>
95 </li>
96 <li>
97 Robert Hegemann
98 <ul>
99 <li>
100 Improved detection of MPEG audio data in RIFF WAVE files.
101 Tracker item <i>[ 3545112 ] Invalid sampling detection</i>
102 </li>
103 <li>
104 New switch <b>--gain &lt;decibel&gt;</b>, range -20.0 to
105 +12.0, a more convenient way to apply Gain adjustment in
106 decibels, than the use of <i>--scale &lt;factor&gt;</i>.
107 </li>
108 <li>
109 Fix for tracker item <i>[ 3558466 ] Bug in path handling</i>
110 </li>
111 <li>
112 Fix for tracker item <i>[ 3567844 ] problem with Tag
113 genre</i>
114 </li>
115 <li>
116 Fix for tracker item <i>[ 3565659 ] no progress
117 indication with pipe input</i>
118 </li>
119 <li>
120 Fix for tracker item <i>[ 3544957 ] scale (empty) silent
121 encode without warning</i>
122 </li>
123 <li>
124 Fix for tracker item <i>[ 3580176 ] environment variable
125 LAMEOPT doesn't work anymore</i>
126 </li>
127 <li>
128 Fix for tracker item <i>[ 3608583 ] input file name
129 displayed with wrong character encoding (on windows
130 console with CP_UTF8)</i>
131 </li>
132 <li>
133 Fix for bug ticket <i>[ #447 ] Fix dereference NULL and
134 Buffer not NULL terminated issues. Thanks to Surabhi
135 Mishra</i>
136 </li>
137 <li>
138 Fix for bug ticket <i>[ #445 ] dereference of a null
139 pointer possible in loop. Thanks to Renu Tyagi</i>
140 </li>
141 <li>
142 Fix for bug ticket <i>[ #449 ] Make sure functions with
143 SSE instructions maintain their own properly aligned
144 stack. Thanks to Fabian Greffrath</i>
145 </li>
146 <li>
147 Fix for bug ticket <i>[ #458 ] Multiple Stack and Heap
148 Corruptions from Malicious File. Thanks to Gareth Evans
149 and Elio Blanca</i>
150 </li>
151 <li>
152 Fix for bug ticket <i>[ #460 ] A division by zero
153 vulnerability. Thanks to Wang Shiyang, Liu Bingchang</i>
154 </li>
155 <li>
156 Fix for bug ticket <i>[ #461 ] CVE-2017-9410
157 fill_buffer_resample function in libmp3lame/util.c
158 heap-based buffer over-read and ap</i>
159 </li>
160 <li>
161 Fix for bug ticket <i>[ #462 ] CVE-2017-9411
162 fill_buffer_resample function in libmp3lame/util.c
163 invalid memory read and application crash</i>
164 </li>
165 <li>
166 Fix for bug ticket <i>[ #463 ] CVE-2017-9412
167 unpack_read_samples function in frontend/get_audio.c
168 invalid memory read and application crash</i>
169 </li>
170 <li>
171 Fix for bug ticket <i>[ #434 ] clip detect scale
172 suggestion unaware of scale input value</i>
173 </li>
174 <li>
175 HIP decoder bug fixed: decoding mixed blocks of lower
176 sample frequency Layer3 data resulted in internal buffer
177 overflow (write). Thanks to Henri Salo</i>
178 </li>
179 </ul>
180 </li>
181 <li>
182 Alexander Leidinger
183 <ul>
184 <li>
185 Feature request, patch ticket <i>[ #27 ] Add
186 lame_encode_buffer_interleaved_int() by Michael Fink</i>
187 </li>
188 </ul>
189 </li>
190 </ul>
191 <br>
192 <hr>
193 <h3>LAME 3.99.5 &nbsp;February 28 2012</h3>
194 <ul>
195 <li>
196 Rog&eacute;rio Brito
197 <ul>
198 <li>
199 autotools: Fix compilation on alpha using proper ifdef
200 guards. Thanks to Andres Mejia.
201 </li>
202 <li>Small correction of the documentation.</li>
203 </ul>
204 </li>
205 <li>
206 Robert Hegemann
207 <ul>
208 <li>
209 Workaround for cygwin build problem. Tracker item
210 <i>[ 3480692 ] mingw-w64 compilation error</i>
211 </li>
212 <li>
213 Fix for a problem when LAME is compiled with <i>fast
214 floating point math</i>. It could result in <i>bitrate
215 bloat</i> with silent input data.
216 </li>
217 <li>
218 <font color="#ff0000">Bug fix for tracker item <i>[
219 3486753 ] Artifacts at the beginning of <b>decoded</b>
220 file</i></font>
221 </li>
222 </ul>
223 </li>
224 </ul>
225 <br>
226 <h3>LAME 3.99.4 &nbsp;January 25 2012</h3>
227 <ul>
228 <li>
229 Robert Hegemann
230 <ul>
231 <li>
232 Fix for tracker item <i>[ 3475581 ] lame crashes at .w64
233 input file</i>
234 </li>
235 <li>
236 Addressing things brought to attention by tracker item
237 <i>[ 3463197 ] 3.99.x problem WFED and PCST frames </i>
238 <ul>
239 <li>
240 WFED and PCST frames can now be added, to tag
241 podcasts iTunes recognizes</li>
242 <li>USER frames are now supported</li>
243 <li>
244 COMM frames can now have a description, when
245 passed via --tv "COMM=description=full text"
246 </li>
247 <li>
248 possible divide-by-zero exception should be fixed
249 </li>
250 <li>
251 adding malformed user-defined-frames could result
252 in abnormal program termination, fixed
253 </li>
254 </ul>
255 </li>
256 </ul>
257 </li>
258 </ul>
259 <br>
260 <h3>LAME 3.99.3 &nbsp;November 26 2011</h3>
261 <ul>
262 <li>
263 Robert Hegemann
264 <ul>
265 <li>
266 Fix for tracker item <i>[ 3441349 ] --tg does not handle
267 genre number when adding unicode tag</i>
268 </li>
269 </ul>
270 </li>
271 </ul>
272 <br>
273 <h3>LAME 3.99.2 &nbsp;November 18 2011</h3>
274 <ul>
275 <li>
276 Robert Hegemann
277 <ul>
278 <li>
279 Due to some bugs in 3rd party HW/SW decoders, those were not
280 be able to make use of LAME's extended info tag anymore,
281 resulting in problems with things like gapless playback.
282 </li>
283 </ul>
284 </li>
285 </ul>
286 <br>
287 <h3>LAME 3.99.1 &nbsp;November 5 2011</h3>
288 <ul>
289 <li>
290 Robert Hegemann
291 <ul>
292 <li>
293 Fixes for several issues with ID3v2 unicode tags, using
294 Big-Endian text encodings. Because of several other
295 software (like Windows Media Player),
296 LAME writes Little-Endian unicode tags only.<br>
297 Thanks to Taihei Monma, for reporting these issues.
298 Tracker items: <i>[ 3431203, 3431222, 3431241 ]</i>
299 </li>
300 </ul>
301 </li>
302 </ul>
303 <br>
304 <h3>LAME 3.99 &nbsp;October 15 2011</h3>
305 <ul>
306 <li>3.99 beta 1 becomes 3.99</li>
307 </ul>
308 <br>
309 <h3>LAME 3.99 beta 1 &nbsp;not yet released</h3>
310 <ul>
311 <li>
312 Robert Hegemann
313 <ul>
314 <li>
315 Fix for Bugtracker item <i>[ 3395813 ]
316 id3tag_set_fieldvalue for URL link frames Broken</i>
317 </li>
318 <li>
319 Fix for Bugtracker item <i>[ 3368977 ] Wave Parser
320 doesn't pad chunks</i>
321 </li>
322 <li>
323 Fix for Bugtracker item <i>[ 3034259 ] confusing fatal
324 error: can't update LAME-tag frame!</i>
325 </li>
326 </ul>
327 </li>
328 </ul>
329 <h3>LAME 3.99 beta 0 &nbsp;not officially released</h3>
330 <ul>
331 <li>
332 Josep Maria Antol&iacute;n Segura
333 <ul>
334 <li>
335 Documentation work
336 </li>
337 </ul>
338 </li>
339 <li>
340 Rog&eacute;rio Brito
341 <ul>
342 <li>
343 Improve Debian packaging.
344 <ul>
345 <li>Include suggestions made by Fabian Greffrath</li>
346 <li>
347 Update version of the Debian package to reflect
348 that it is taken from CVS.
349 </li>
350 <li>Get rid of a transitional package</li>
351 <li>Much more to come...</li>
352 </ul>
353 <li>Minor fix to validate the history file correctly</li>
354 </ul>
355 </li>
356 <li>
357 Robert Hegemann
358 <ul>
359 <li>
360 Porting of LAME <b>mp3rtp</b> program to Windows platform
361 </li>
362 <li>
363 Fix: <font color="#ff0000">for free format mp3 above 320
364 kbps, the maximum allowed bits was limited too far.</font>
365 </li>
366 <li>
367 Changed behaviour: LAME replaces common suffixes by
368 MP3/WAV when no output filename was given. For example,
369 a simple <tt>lame x.wav</tt> will create output file
370 <tt>x.mp3</tt>
371 </li>
372 <li>
373 UNICODE support on Windows OS, file names and ID3v2 tags;
374 related: Bugtracker item <i>[ 3052230 ] Lame.exe can't
375 find files with non-ASCII names</i>
376 </li>
377 <li>
378 Added support for WAVE FORMAT IEEE FLOAT input files
379 </li>
380 <li>
381 New switch <b>--swap-channel</b> added, see tracker
382 item <i>[ 1118412 ] RFE: option to swap L and R channels</i>
383 </li>
384 <li>
385 New switches <b>-m l</b> and <b>-m r</b> to select either
386 left or right channel for mono encoding.
387 </li>
388 <li>
389 Feature request item <i>[ 3134258 ] Allow larger images
390 with LAME (>128KB)</i>
391 </li>
392 <li>
393 Fix for Bugtracker item <i>[ 2962223 ] Encoder delay
394 and padding values ignored when re-encoding</i>
395 </li>
396 <li>
397 Fix for Bugtracker item <i>[ 2986823 ] Error when
398 building Lame with NASM support -still exists</i>
399 </li>
400 <li>
401 Fix for Bugtracker item <i>[ 3125235 ] make -f
402 Makefile.mingw32 fails</i>
403 </li>
404 <li>
405 Fix for Bugtracker item <i>[ 3021935 ] ID3 tag TXXX
406 cannot be used more than once</i>
407 </li>
408 <li>
409 Fix for Bugtracker item <i>[ 3025801 ] gain_analysis.h
410 needs int constants</i>
411 </li>
412 <li>
413 Fix for Bugtracker item <i>[ 3277412 ] Segmentation
414 fault at layer2.c:105 with --decode</i>
415 </li>
416 <li>
417 Changes in new VBR code:
418 <ul>
419 <li><font color="#ff0000">tuning on PSY model</font></li>
420 <li>
421 <font color="#770000">tuning on VBR scale and
422 resulting bitrates</font>
423 </li>
424 <li>
425 <font color="#ff0000">Minor bug fix for sfb21
426 encoding</font>
427 </li>
428 </ul>
429 </li>
430 <li>
431 <font color="#ff0000">All encoding modes use the PSY
432 model from new VBR code, addresses Bugtracker item <i>[
433 3187397 ] Strange compression behavior</i></font>
434 </li>
435 <li>
436 Fix for a possible out of bounds array access in HIP
437 decoder, addresses Bugtracker item <i>[ 3091776 ]
438 crash at III_dequantize_sample</i>
439 </li>
440 </ul>
441 </li>
442 </ul>
443 <hr>
444 <h3>LAME 3.98.4 &nbsp;March 22 2010</h3>
445 <ul>
446 <li>
447 Joseph Flynn
448 <ul>
449 <li>
450 Improvements for LAME DirectShow filter:
451 <ul>
452 <li>Added support for the DirectShow IAMStreamConfig Interface to the LAME encoder filter output pin.
453 </li>
454 <li>Modified the DirectShow filter registration section so that the LAME Encoder filter is correctly registered in the Audio Compressors filter category. This will allow third-party encoding applications using the DirectShow System Device Enumerator Interface to correctly detect the LAME encoder when querying the Audio Compressors filter group.
455 </li>
456 <li>Modified the filter registration information so that the MP3 audio subtype is correctly reported as being supported on the encoder output pin. This will allow third-party encoding applications using the DirectShow IFilterMapper2 Interface to recognize that the LAME encoder supports MP3 output.
457 </li>
458 <li>Altered the Filter Merit Value that was being used when the filter was registered so that it is now using the standard DirectShow compressor filter merit value of MERIT_DO_NOT_USE (0x200000). Previously, the filter was being registered using a value of MERIT_SW_COMPRESSOR (0x100000), which was at a lower priority (i.e. worse priority) than MERIT_DO_NOT_USE. This prevented the LAME Encoder filter from being selected for use by some third-party encoding applications.
459 </li>
460 <li>Added code to calculate the frame length of the audio frames used for the nBlockSize element of the WAVEFORMATEX output structure. Previously this value was simply hard-coded to 1.
461 </li>
462 </ul>
463 </li>
464 </ul>
465 </li>
466 <li>
467 Robert Hegemann
468 <ul>
469 <li><font color="#ff0000">Fix for Bugtracker item <i>[ 2973877 ] A problem regarding the new drain code</i></font>
470 </li>
471 </ul>
472 </li>
473 </ul>
474 <h3>LAME 3.98.3 &nbsp;February 27 2010</h3>
475 <ul>
476 <li>
477 Rog&eacute;rio Brito:
478 <ul>
479 <li>Update the debian packaging for the new release.
480 </li>
481 </ul>
482 </li>
483 <li>
484 Robert Hegemann
485 <ul>
486 <li>The <b>ignore-tag-errors</b> switch had no effect when embedding album art, fixed.</li>
487 <li>Library API change: lame_decode functions are now obsolete but still present, please use hip_decode instead.
488 The reason for this change is: lame_decode functions use a single global variable within the library
489 to store decoder setup, hip_decode functions don't.
490 The encoder now uses hip_decode internally and it is now possible to use <b>clipdetect</b>
491 feature while reencoding mp3 to mp3.
492 </li>
493 <li>Workaround for FFMPEG bug, which uses to call lame_encode_flush more than once in a loop.
494 </li>
495 <li>Windows: program icon and version info added (when building with VC9)
496 </li>
497 <li>Fix for Bugtracker item <i>[ 2688413 ] lib name problem in Microsoft Visual Studio 6</i>
498 </li>
499 <li>Fix for Bugtracker items <i>[ 2051870, 2423650, 2928684 ] several small documentation issues</i>
500 </li>
501 <li>Fix for Bugtracker item <i>[ 2723518 ] resampling in 3.98 and 3.99alpha</i>
502 </li>
503 <li>Fix for Bugtracker item <i>[ 2891879 ] Because of Windows API change, there was a problem with the <b>prority</b> switch.</i>
504 </li>
505 <li>Fix for Bugtracker item <i>[ 2893101 ] Access Violation in BladeMP3EncDLL if UNICODE was defined.</i>
506 </li>
507 <li>Fix for Bugtracker item <i>[ 2887359 ] Wrong length in ID3v2 tag when num_samples isn't set</i>
508 </li>
509 <li>Fix for Bugtracker item <i>[ 2872590 ] LameTAG: "Music length" missmatch in LAME 3.98</i>
510 </li>
511 <li>Fix for Bugtracker item <i>[ 2824296 ] wrong enc_padding value in LAME 3.99a and 3.98.3 (from CVS)</i>
512 </li>
513 <li><font color="#ff0000">Revisiting the FhG decoder problem (FhG V1.5 build 50, ships with MS Windows):
514 enabling the new-drain-code seems to solve that issue better, than restricting the buffer size (see below: 3.98 beta 1, May 16 2007).</font>
515 </li>
516 <li>Patch submitted by Bernhard Doebler, tracker item <i>[ 2807676 ] Error when building Lame with NASM support</i>
517 </li>
518 <li>Patch submitted by Mancuso Raffaele, tracker item <i>[ 2406420 ] compile lame_enc.dll under cygwin</i>
519 </li>
520 </ul>
521 </li>
522 </ul>
523 <h3>LAME 3.98.2 &nbsp;September 22 2008</h3>
524 <ul>
525 <li>
526 Robert Hegemann
527 <ul>
528 <li>Fix for Bugtracker item <i>[ 2123206 ] lame 3.98.1 segfaults with -h</i>
529 </li>
530 </ul>
531 </li>
532 </ul>
533 <h3>LAME 3.98.1 &nbsp;September 21 2008</h3>
534 <ul>
535 <li>
536 Rog&eacute;rio Brito:
537 <ul>
538 <li>
539 More fixes for the abx tool for Unix systems:
540 <ul>
541 <li>Plugged a memory leak.</li>
542 <li>Fixed an endianness problem: users of big-endian machines
543 can now do abx tests.
544 </li>
545 </ul>
546 </li>
547 <li>Fixed history's HTML doctype</li>
548 <li>
549 Fixed history so that it <em>finally</em> validates
550 at <a href="http://validator.w3.org/">W3's validator</a>
551 </li>
552 <li>
553 Fixed compilation of frontend <code>mp3rtp.c</code>. Thanks to Kris Karas.
554 Bugtracker item <i>[ 2015432 ] mp3rtp missing uint16_t in lame 3.98</i>
555 </li>
556 </ul>
557 </li>
558 <li>
559 Robert Hegemann:
560 <ul>
561 <li>Fix for Bugtracker item <i>[ 2031704 ] --id3v1-only didnt work in 3.98-final</i></li>
562 <li>Fix for Bugtracker item <i>[ 2022035 ] encoder_padding value and resampling</i></li>
563 <li>Fix for Bugtracker item <i>[ 2029282 ] Frequency filtering API broken in 3.98</i></li>
564 <li>Fix for Bugtracker item <i>[ 2039648 ] potential memory leak in parse_args() function in parse.c</i></li>
565 <li>
566 Fix for some tagging issues:
567 <ul>
568 <li>
569 Made search for ID3v1 genres more sloppy, abbrevations may match more often as some simple typos.
570 Examples:
571 <ul>
572 <li>--tg "Alt. Rock" matches genre "Alternate Rock"</li>
573 <li>--tg "acapela" matches genre "A Cappella"</li>
574 </ul>
575 </li>
576 <li>New switch --pad-id3v2-size "n": adds ID3v2 tag with n padding bytes.</li>
577 </ul>
578 </li>
579 </ul>
580 </li>
581 </ul>
582 <hr>
583 <h3>LAME 3.98 &nbsp;July 4 2008</h3>
584 <ul>
585 <li>
586 Anton Sergunov:
587 <ul>
588 <li>Frontend DirectShow: enabling LAME dshow filter to connect to "File Writer Filter".
589 </li>
590 </ul>
591 </li>
592 <li>
593 Rog&eacute;rio Brito:
594 <ul>
595 <li>Updates to the Debian Packaging</li>
596 <li>Fixes to the abx tool for Unix systems (so that more people
597 can evaluate LAME's compression against the original files)
598 </li>
599 </ul>
600 </li>
601 <li>
602 Alexander Leidinger:
603 <ul>
604 <li>explicitely link the math lib to the lame lib</li>
605 <li>add switch to disable the use of the compaq optimized math lib</li>
606 </ul>
607 </li>
608 </ul>
609 <h3>LAME 3.98 beta 8 &nbsp;April 13 2008</h3>
610 <ul>
611 <li>
612 Robert Hegemann:
613 <ul>
614 <li>LAME now accepts a floating point value in the range [0,...,10[ as VBR quality setting, like <b>-V5.678</b>
615 </li>
616 <li>Found and fixed some suspicious code in additive masking calculation for VBR-NEW
617 </li>
618 <li>bug-fix:<font color="#ff0000">experimental code was defaulted by accident for VBR-NEW</font>
619 </li>
620 <li>fix for some endianess problem on big-endian machines
621 </li>
622 </ul>
623 </li>
624 </ul>
625 <h3>LAME 3.98 beta 7 &nbsp;April 6 2008</h3>
626 <ul>
627 <li>
628 Robert Hegemann:
629 <ul>
630 <li>libmp3lame API: allow frontends to separately retrieve LAME/Xing and ID3 data, because the old library automatism
631 makes it impossible to make fully buffered encodes.
632 </li>
633 <li>libmp3lame API: added some experimental unicode ID3 tagging code.
634 </li>
635 <li>frontends: write itself final ID3 tags and LAME/Xing header frame
636 </li>
637 <li>lame_enc.dll: writes itself final LAME/Xing header frame
638 </li>
639 <li>
640 Latest changes to the new VBR psymodel:
641 <ul>
642 <li>uses a different spreading function
643 </li>
644 <li><font color="#ff0000">bug-fix for out-of-bounds array access (program stack corruption possible)</font>
645 </li>
646 </ul>
647 </li>
648 </ul>
649 </li>
650 </ul>
651 <h3>LAME 3.98 beta 6 &nbsp;December 16 2007</h3>
652 <ul>
653 <li>
654 Robert Hegemann:
655 <ul>
656 <li>Feature request <i>[ 1811483 ] WAVE_FORMAT_EXTENSIBLE support (PCM)</i>
657 </li>
658 <li><font color="#ff0000">Fix for some rare scalefactor selection issue the newer vbr code had at low compression levels
659 </font>
660 </li>
661 <li>Fix for Bugtracker item<i>[ 1813496 ] AIFF parsing bug</i>
662 </li>
663 <li>
664 Latest changes to the new VBR code:
665 <ul>
666 <li><font color="#3366ff">it now has its own psy model, a derivation from NSPSY.</font>
667 </li>
668 <li><font color="#ff0000">some more tuning has been done for this new psy model.</font>
669 Many thanks to Horst Albrecht and Myles Thaiss.
670 </li>
671 <li><font color="#ff0000">the "out-of-bits" strategy is reworked</font>
672 </li>
673 </ul>
674 <li><font color="#ff0000">It was possible, that the "interchannel masking effects feature"
675 was used by the dual-channel-mode for bi-lingual encodings too. It was meant to work
676 on stereo L/R channels only.</font>
677 </li>
678 </ul>
679 </li>
680 </ul>
681 <h3>LAME 3.98 beta 5 &nbsp;August 12 2007</h3>
682 <ul>
683 <li>
684 Jonathan Stott:
685 <ul>
686 <li>Bug tracker items: <i>[ 1590693 ] ID3v2 tag not writing, [ 1636267 ] ID3v2 tags overwritten</i><br>
687 If the output file is opened 'write-only', then LAME can't update the LAME tag.
688 In this case LAME silently overwrote the first bytes of the file and an
689 optional ID3v2 tag disappeared. Now an error message will be printed and
690 no data is written in this case.
691 </li>
692 </ul>
693 </li>
694 <li>
695 Robert Hegemann:
696 <ul>
697 <li>Fix for Bugtracker item <i>[ 1719593 ] Track numbers > 255 not allowed even with --id3v2-only</i>
698 <li>Fix for Bugtracker item <i>[ 1742623 ] fail(lame --mp3input -m m -b 128 --resample 8 *.mp3 **.mp3)</i><br>
699 The problem here was, the input files are MPEG-1 Layer2 files named as MP3s. Even if you leave out
700 the --mp3input switch LAME tried to decode the input files as Layer3 files because of the file name
701 extension and because it found some valid looking Layer3 synchronization header.
702 The fixed LAME version does not assume the file name extension is always correct and treats the files
703 depending on the first found MPEG sync word. The files in question are now correctly detected as
704 Layer2 files and transcoding does succeed.
705 </li>
706 <li>Fix for Bugtracker item <i>[ 1445175 ] Input being stdin fails in Windows on WAV files</i><br>
707 The problem here was, seeking on pipes shows some different behaviour depending on C-Library
708 implementations. The workaround tries to detect it's working on a pipe and doing some
709 reading instead of seeking in that case.
710 </li>
711 <li>Fixing some memory leak in the 'lame_enc.dll'.
712 </li>
713 <li>Fix for Bugtracker items <i>[ 1160757, 1160741 ] --little-endian / --big-endian not working</i><br>
714 These switches where originally intended to be used together with Libsndfile only.
715 </li>
716 <li>Fix for Bugtracker item <i>[ 1746336 ] Incorrect Bitrate with ABR und --resample, LAME 3.98b4</i><br>
717 Some earlier bug-fix had some typo. As a result, when adding a '--resample 123' switch,
718 the average bitrate rised upto maximum bitrate.
719 </li>
720 </ul>
721 </li>
722 </ul>
723 <h3>LAME 3.98 beta 4 &nbsp;June 23 2007</h3>
724 <ul>
725 <li>Dennis Lambe Jr: Added support for total track count (id3v2) in the frontend
726 </li>
727 <li>
728 Nyaochi:
729 <ul>
730 <li>Ability to set user-defined ID3v2.3 frame</li>
731 <li>Ability to include albumArt in ID3v2.3 tag</li>
732 </ul>
733 </li>
734 <li>
735 Robert Hegemann:
736 <ul>
737 <li>Bugfix: the "play length in ms", which is stored in the ID3v2 tag TLEN, was not correctly computed.
738 Some hardware and software players were confused by this garbage data.
739 </li>
740 <li><font color="#ff0000">Out of bits strategy for the newer VBR code overhauled</font></li>
741 <li>LAME API: the ID3 tag functions do not store the pointers passed anymore,
742 they do make deep copies of strings passed as parameters.
743 </li>
744 <li>Changes in LAME frontend switches regarding ID3 tags:
745 <br><tt>--tg "MyGenre"</tt> will route unknown ID3v1 genres to "Other" for ID3v1 tags
746 and will be stored as plain text "MyGenre" for ID3v2 tags. Genres given by known
747 ID3v1 numbers will be stored as its corresponding text in ID3v2 tags.
748 <br><tt>--tn "02/02"</tt> will store the track number specified as plain text as-is
749 for ID3v2 tags.
750 </li>
751 </ul>
752 </li>
753 </ul>
754 <br>
755 <h3>LAME 3.98 beta 3 &nbsp;May 22 2007</h3>
756 <ul>
757 <li>
758 Robert Hegemann:
759 <ul>
760 <li>Fixes regarding max number of bits limitation</li>
761 </ul>
762 </li>
763 </ul>
764 <br>
765 <h3>LAME 3.98 beta 2 &nbsp;May 20 2007</h3>
766 <ul>
767 <li>
768 Robert Hegemann:
769 <ul>
770 <li>Bug tracker item: <i>[ 1693461 ];</i>
771 Fixed memory leaks in ACM codec
772 </li>
773 <li>Fixed encoding of non-standard sampling rates in CBR</li>
774 <li><font color="#ff0000">Improved VBR strategy when running out of bits</font></li>
775 </ul>
776 </li>
777 </ul>
778 <br>
779 <h3>LAME 3.98 beta 1 &nbsp;May 16 2007</h3>
780 <ul>
781 <li>
782 Alexander Leidinger:
783 <ul>
784 <li>Add TLEN (ID3v2) support (Submitted by: Linus Walleij).</li>
785 <li>Add number of total tracks per album (ID3v2) support
786 (Submitted by: Kyle VanderBeek).
787 </li>
788 <li>Some seatbelts for overflowing arrays in the ID3v2
789 support.
790 </li>
791 <li>Update the RPM spec (Submitted by: Kyle VanderBeek).</li>
792 <li>Fix some mem-leaks in the error case.</li>
793 <li>Update to newer autotools versions.</li>
794 <li>Update to use a recent libsndfile (submitted by
795 libsndfile author).
796 </li>
797 <li><font color="#3366ff">Intrinsics support
798 enabled for gcc</font>
799 </li>
800 </ul>
801 </li>
802 <li>
803 Robert Hegemann:
804 <ul>
805 <li><font color="#3366ff">The newer VBR code is now LAME's default
806 VBR routine</font>
807 </li>
808 <li><font color="#ff0000">Fixed: in case of
809 not enough bits the new vbr code incorrectly used old vbr routine</font>
810 </li>
811 <li><font color="#ff0000">Improved ATH
812 adjustment in low volume cases</font>
813 </li>
814 <li><font color="#ff0000">Fixed (PSY model): mapping from convolution bands
815 to partition bands was broken since we replaced tables by own calculation
816 several years ago</font>
817 </li>
818 <li><font color="#ff0000">Fixed (PSY model): loss of fraction in equal loudness weighting</font></li>
819 <li><font color="#ff0000">Fixed (PSY model): in NSPSY highpass filter, out of bounds access in fircoef</font></li>
820 <li><font color="#ff0000">Known problem samples for the new VBR code:
821 many of them are at an acceptable quality level now;</font>
822 with a big <b>'Thank You'</b> to Francis Niechcial
823 </li>
824 <li><font color="#ff0000">Modified VBR strategy to handle out of bits cases</font></li>
825 <li>Restricted bitreservoir size for 320 kbps frames to
826 the size used for sideinfo, because of decoding problems
827 with FhG decoders installed on almost every Windows system
828 </li>
829 <li>LAME aborts on unsupported input files
830 or unrecognized parameter options passed more often now
831 </li>
832 <li>Bug tracker item: <i>[ 1596306 ] "fatal error during initialization";</i>
833 an invalid MPEG samplerate was returned by optimum_samplefreq function
834 </li>
835 <li>Bug tracker item: <i>[ 1585942 ] lame not --silent when TERM not set;</i>
836 in case LAME was build with TERMCAP defined and no TERM
837 environment is defined, now we do not issue an error message and
838 silently fallback to the default behaviour as if LAME was
839 compiled without TERMCAP defined.
840 </li>
841 <li>Bug tracker item: <i>[ 1711980 ] LAME writes invalid Xing header when ID3 tags exist;</i>
842 LAME was sometimes writing an invalid Xing/Info header
843 </li>
844 <li>Feature request: <i>[ 1588283 ] Flushing output stream in lame.exe;</i>
845 'flush' option added
846 </li>
847 <li><font color="#3366ff">Added FFTSSE and FFT3DNOW assembler code from
848 Lame4 branch</font>
849 </li>
850 <li>Changes in lame frontend switches:
851 -k removed, add lowpass and highpass switches if you need to change
852 them; --short/noshort/allshort - degraded into DEVELOPER ONLY switches
853 normal users shouldn't use them; -X -Z degraded to
854 DEVELOPER ONLY switches, -X is too tough to communicate to
855 end users and -Z isn't used actualy
856 </li>
857 <li>Fixed some console printing problems</li>
858 <li>Windows: ACM code now uses LAME library API only, all
859 references to private include files are removed
860 </li>
861 <li>Windows: DirectShow code now uses LAME library API only, all
862 references to private include files are removed
863 </li>
864 <li>Windows: disabled code that resets processor affinity,
865 because this doesn't belong to LAME, but seems to work around
866 some problems the parent process has (in most cases EAC)
867 </li>
868 </ul>
869 </li>
870 <li>
871 John33:
872 <ul>
873 <li>
874 Fixed mp2 and mp3 decoding: For mp3 and mp2 decoding,
875 this now yields the same output as foobar2000 but the
876 error checking remains unchanged
877 </li>
878 </ul>
879 </li>
880 <li>
881 Gabriel Bouvigne:
882 <ul>
883 <li>VC8 project files</li>
884 <li>Added support for x64 under VC8</li>
885 <li>Restricted MPEG 2.5 to 64kbps frames</li>
886 </ul>
887 </li>
888 <li>
889 Takehiro Tominaga:
890 <ul>
891 <li><font color="#3366ff">SSE version of FFT</font></li>
892 </ul>
893 </li>
894 </ul>
895 <br>
896 <h3>LAME 3.97 &nbsp;September 24 2006</h3>
897 <ul>
898 <li>3.97 beta 3 becomes 3.97 </li>
899 </ul>
900 <br>
901 <h3>LAME 3.97 beta 3 &nbsp;August 19 2006</h3>
902 <ul>
903 <li>
904 Gabriel Bouvigne:
905 <ul>
906 <li><font color="#ff0000">Workaround against a
907 short blocks detection issue</font>
908 </li>
909 </ul>
910 </li>
911 </ul>
912 <br>
913 <h3>LAME 3.97 beta 2 &nbsp;November 26 2005</h3>
914 <ul>
915 <li>
916 Gabriel Bouvigne:
917 <ul>
918 <li>Fixed an initialization error when input is not using a
919 standard sampling frequency
920 </li>
921 <li>Fixed a possible assertion failure in very low bitrate
922 encoding
923 </li>
924 <li><font color="#ff0000">Slight change
925 regarding ATH adjustment with V5</font>
926 </li>
927 <li><font color="#ff0000">Reinstated bit
928 reservoir for 320kbps CBR</font>
929 </li>
930 <li><font color="#3366ff">ReplayGain analysis
931 should now be faster when encountering silent parts</font>
932 </li>
933 </ul>
934 </li>
935 <li>
936 Takehiro Tominaga:
937 <ul>
938 <li>Fixed a possible link problem of assembly code</li>
939 </ul>
940 </li>
941 </ul>
942 <br>
943 <h3>LAME 3.97 beta 1 &nbsp;September 12 2005</h3>
944 <ul>
945 <li>
946 Takehiro Tominaga:
947 <ul>
948 <li>Fixed an out of array access in mp3rtp </li>
949 <li><font color="#ff0000">Fixed a quality
950 setting in DLL</font>
951 </li>
952 <li>Fixed display when using --silent</li>
953 </ul>
954 </li>
955 <li>
956 Vitaly Ivanov:
957 <ul>
958 <li>Updated DirectShow interface</li>
959 </ul>
960 </li>
961 <li>
962 Robert Hegemann:
963 <ul>
964 <li><font color="#ff0000">Fixed an out of
965 array access</font>
966 </li>
967 <li><font color="#ff0000">Fixed some small
968 rounding problem in vbr-new quantization routines</font>
969 </li>
970 <li><font color="#ff0000">Fixed a bug in
971 vbr-new regarding high frequencies (sfb21) when using -Y</font>
972 </li>
973 <li><font color="#ff0000">Fixed a few bugs in
974 vbr-new when using -Y</font>
975 </li>
976 <li><font color="#ff0000">Updated scalefactors
977 allocation scheme in vbr-new</font>
978 </li>
979 <li>Fixed mingw32 configure problems</li>
980 <li>Resolved some compiler warnings</li>
981 <li>Updated command-line visualisation</li>
982 </ul>
983 </li>
984 <li>
985 Gabriel Bouvigne:
986 <ul>
987 <li>Changed some FLOAT8 to FLOAT</li>
988 <li>Added project files for VC7</li>
989 <li><font color="#ff0000">Reworked -q1 and -q0</font></li>
990 <li><font color="#ff0000">Updated presets</font></li>
991 <li><font color="#ff0000">Fixed an error in
992 ISO quantization on systems not using the IEEE754 hack</font>
993 </li>
994 <li><font color="#3366ff">Faster quantization</font></li>
995 <li><font color="#3366ff">SSE version of
996 init_xrpow</font>
997 </li>
998 </ul>
999 </li>
1000 <li>
1001 Rog&eacute;rio Brito:
1002 <ul>
1003 <li>Updated Debian packaging</li>
1004 <li>Documentation work</li>
1005 </ul>
1006 </li>
1007 <li>
1008 Chris Miller:
1009 <ul>
1010 <li>Support for x64 platform SDK in makefile.msvc</li>
1011 </ul>
1012 </li>
1013 </ul>
1014 <br>
1015 <h3>LAME 3.96.1 &nbsp;July 25 2004</h3>
1016 <ul>
1017 <li>
1018 Robert Hegemann:
1019 <ul>
1020 <li><font color="#ff0000">Fixed a rare bug in
1021 vbr-new (could lead to crashes or data corruption)</font>
1022 </li>
1023 </ul>
1024 </li>
1025 <li>
1026 Gabriel Bouvigne:
1027 <ul>
1028 <li>some fixes in ACM codec</li>
1029 <li>fixed padding when encoding to 320kbps</li>
1030 <li><font color="#ff0000">fixed block size
1031 selection for mid and side channels</font>
1032 </li>
1033 </ul>
1034 </li>
1035 </ul>
1036 <br>
1037 <h3>LAME 3.96 &nbsp;April 11 2004</h3>
1038 <ul>
1039 <li>
1040 Gabriel Bouvigne:
1041 <ul>
1042 <li><font color="#ff0000">new quantization
1043 selection mode (used in ABR/CBR)</font>
1044 </li>
1045 <li><font color="#ff0000">set sfscale for
1046 ABR/CBR up to 160kbps</font>
1047 </li>
1048 </ul>
1049 </li>
1050 </ul>
1051 <br>
1052 <h3>LAME 3.96 beta 2 &nbsp;March 28 2004</h3>
1053 <ul>
1054 <li>
1055 Takehiro Tominaga:
1056 <ul>
1057 <li><font color="#3366ff">removed unnecessary
1058 integer convertion in resampling</font>
1059 </li>
1060 </ul>
1061 </li>
1062 <li>
1063 Robert Hegemann:
1064 <ul>
1065 <li><font color="#ff0000">reworked scalefactor
1066 allocation in vbr-new</font>
1067 </li>
1068 <li>fixed a freeformat decoding problem</li>
1069 </ul>
1070 </li>
1071 <li>
1072 Gabriel Bouvigne:
1073 <ul>
1074 <li><font color="#ff0000">updated minimal
1075 bitrate for V1 and V2</font>
1076 </li>
1077 </ul>
1078 </li>
1079 <li>
1080 Aleksander Korzynski:
1081 <ul>
1082 <li>added ability to disable ReplayGain analysis</li>
1083 </ul>
1084 </li>
1085 </ul>
1086 <br>
1087 <h3>LAME 3.96 beta &nbsp;March 7 2004</h3>
1088 <ul>
1089 <li>
1090 Takehiro Tominaga:
1091 <ul>
1092 <li>fixed decoding issue</li>
1093 </ul>
1094 </li>
1095 <li>
1096 Aleksander Korzynski:
1097 <ul>
1098 <li>changed internal ReplayGain handling</li>
1099 <li>fixed some issues when ReplayGain is used with
1100 resampling
1101 </li>
1102 </ul>
1103 </li>
1104 <li>
1105 Robert Hegemann:
1106 <ul>
1107 <li>added standard ISO quantization for vbr-new, used at
1108 lower quality settings
1109 </li>
1110 <li><font color="#3366ff">faster count_bits
1111 for vbr-new</font>
1112 </li>
1113 <li><font color="#3366ff">faster
1114 find_scalefac_ave function for vbr-new</font>
1115 </li>
1116 <li><font color="#ff0000">fixed an out of
1117 array access in psychoacoustic models; this bug could make some psy
1118 calculations worthless and sometimes let lame crash</font>
1119 </li>
1120 <li><font color="#ff0000">fixed an error on
1121 silent scalefactor bands; this bug resulted in huffman data overrun
1122 problems while decoding, resulting in audible glitches</font>
1123 </li>
1124 <li>fixed a freeformat decoding bug</li>
1125 </ul>
1126 </li>
1127 <li>
1128 Gabriel Bouvigne:
1129 <ul>
1130 <li><font color="#ff0000">adjusted short block
1131 thresholds</font>
1132 </li>
1133 <li>fixed some array addressing bugs</li>
1134 <li>made ReplayGain analysis reentrant</li>
1135 </ul>
1136 </li>
1137 <li>David Chandler: fixed a crash in quantize_xrpow </li>
1138 <li>Michal Bacik: fixed a crash when using 8kHz </li>
1139 <li>Goran Markovic: fixed some decoding bugs </li>
1140 <li>John Edwards: fixed a too small buffer in ReplayGain code</li>
1141 </ul>
1142 <br>
1143 <h3>LAME 3.95.1 &nbsp;January 12 2004</h3>
1144 <ul>
1145 <li>
1146 Gabriel Bouvigne:
1147 <ul>
1148 <li>fixed a crash when using vbr-new</li>
1149 <li>changed ReplayGain reference level to 89dB</li>
1150 </ul>
1151 </li>
1152 </ul>
1153 <br>
1154 <h3>LAME 3.95 &nbsp;January 11 2004</h3>
1155 <ul>
1156 <li>
1157 Gabriel Bouvigne:
1158 <ul>
1159 <li><font color="#ff0000">fixed lowpass values
1160 when using vbr with mono files</font>
1161 </li>
1162 <li><font color="#3366ff">faster quantization
1163 loops</font>
1164 </li>
1165 <li><font color="#3366ff">faster count_bits</font></li>
1166 <li>fixed a buffer requirement error in ACM codec</li>
1167 </ul>
1168 </li>
1169 <li>
1170 Takehiro TOMINAGA:
1171 <ul>
1172 <li>fixed mpglib and other decoding support code to prevent
1173 the crash when invalid mp3 input
1174 </li>
1175 </ul>
1176 </li>
1177 <li>removed Layer I decoding support</li>
1178 <li><font color="#3366ff">use FastLog and IEEE 754
1179 hack on PowerPC too (approx. 10 percent faster)</font>
1180 </li>
1181 </ul>
1182 <br>
1183 <h3>LAME 3.94 beta December 15 2003</h3>
1184 <ul>
1185 <li>
1186 Takehiro Tominaga:
1187 <ul>
1188 <li><font color="#ff0000">fixed block
1189 switching of nspsytune</font>
1190 </li>
1191 <li><font color="#ff0000">best huffman divide
1192 in the inner loop.</font> This should improve the quality, but
1193 PAINFULLY slow. So it is not enabled by default. Use -q0 to use it.
1194 </li>
1195 <li>Changed -q option mapping. "-q2" until version 3.93 is
1196 now "-q3".
1197 </li>
1198 <li><font color="#ff0000">saving bits by
1199 better scalefactor storing</font>
1200 </li>
1201 <li>removed Vorbis support</li>
1202 <li><font color="#ff0000">substep quantization.</font>This
1203 should help breaking the SFB21 bloating problem
1204 </li>
1205 <li><font color="#ff0000">made psychoacoustic
1206 model aware of ATH adjustements</font>
1207 </li>
1208 <li><font color="#ff0000">use ATH value as
1209 short block masking lower limit</font>
1210 </li>
1211 <li><font color="#ff0000">several fixes in
1212 psychoacoustic model</font>
1213 </li>
1214 <li>more robust decoding</li>
1215 </ul>
1216 </li>
1217 <li>Mark Taylor / Gabriel Bouvigne: fixed issues in VBR header</li>
1218 <li>Mark Taylor: workaround against some hardware decoder
1219 defficiencies
1220 </li>
1221 <li>Aleksander Korzynski: ability to compute the "Radio"
1222 ReplayGain and detect clipping on the fly. The ReplayGain value is
1223 stored in the Lame tag.
1224 </li>
1225 <li>
1226 Gabriel Bouvigne:
1227 <ul>
1228 <li><font color="#ff0000">work on presets</font></li>
1229 <li><font color="#ff0000">use presets by
1230 default for cbr/abr</font>
1231 </li>
1232 <li><font color="#ff0000">use presets by
1233 default for vbr</font>
1234 </li>
1235 <li><font color="#ff0000">analog silence
1236 detection in partitionned sfb21</font>
1237 </li>
1238 <li><font color="#3366ff">do not compute noise
1239 in upper 0 part of the spectrum</font>
1240 </li>
1241 <li><font color="#3366ff">only compute noise
1242 in modified scalefactor bands</font>
1243 </li>
1244 </ul>
1245 </li>
1246 <li>
1247 Guillaume Lessard:
1248 <ul>
1249 <li>nogap related changes</li>
1250 </ul>
1251 </li>
1252 <li>
1253 Alexander Leidinger:
1254 <ul>
1255 <li>prevent closing the input fd prematurely if the input
1256 is a named pipe
1257 </li>
1258 </ul>
1259 </li>
1260 </ul>
1261 <br>
1262 <h3>LAME 3.93.1 &nbsp;December 1 2002</h3>
1263 <ul>
1264 <li>
1265 Gabriel Bouvigne:
1266 <ul>
1267 <li>preset medium added to the dll interface</li>
1268 <li><font color="#ff0000">fix for abr/cbr
1269 presets</font>
1270 </li>
1271 <li><font color="#ff0000">fix -q0 switch</font></li>
1272 </ul>
1273 </li>
1274 <li>Alexander Leidinger: fix link problem on systems where
1275 socket() resides in libsocket
1276 </li>
1277 </ul>
1278 <br>
1279 <h3>LAME 3.93 &nbsp;November 16 2002</h3>
1280 <ul>
1281 <li>
1282 Takehiro Tominaga:
1283 <ul>
1284 <li><font color="#ff0000">bit allocation for
1285 pre-echo control improved for single channel encodings</font>
1286 </li>
1287 <li><font color="#ff0000">substep noise shaping</font></li>
1288 <li><font color="#3366ff">optimizations by
1289 changing data structure</font>
1290 </li>
1291 <li><font color="#ff0000">noise shaping model
1292 2 fix</font>
1293 </li>
1294 <li><font color="#3366ff">nspsytune FIR filter
1295 clean up</font>
1296 </li>
1297 <li><font color="#ff0000">fix small psymodel
1298 bugs(DC current estimation, preecho detection of non-VBR mode, and
1299 nspsymode initialization)</font>
1300 </li>
1301 <li>portability fixes for Tru64 UNIX</li>
1302 </ul>
1303 </li>
1304 <li>Albert Faber: some fixes in the DLL</li>
1305 <li>Simon Blandford: fixes for channel scaling in mono mode</li>
1306 <li><font color="#3366ff">Dominique Duvivier: some
1307 optimizations and a faster log10 function</font>
1308 </li>
1309 <li>
1310 Mark Taylor:
1311 <ul>
1312 <li>some tag related fixes in the direct show filter and in
1313 the ACM codec
1314 </li>
1315 <li><font color="#3366ff">fixed a mono
1316 encoding bug found by Justin Schoeman</font>
1317 </li>
1318 <li>calc_noise bug fix</li>
1319 <li>other fixes</li>
1320 </ul>
1321 </li>
1322 <li>
1323 Alexander Leidinger:
1324 <ul>
1325 <li>update to autoconf 2.53, rewrite some configure tests</li>
1326 <li>Akos Maroy: determine gcc version even with gcc 3.1</li>
1327 <li>Andrew Bachmann: compile shared libs on BeOS (and
1328 perhaps other arches)
1329 </li>
1330 <li>ultrasparc switches for gcc 3.1</li>
1331 <li>fixes for SunOS 4.x</li>
1332 <li>fixes for 64bit arches</li>
1333 <li>CFLAGS fix for IRIX</li>
1334 <li>don't override CFLAGS if exptopt isn't requested</li>
1335 </ul>
1336 </li>
1337 <li>
1338 Robert Hegeman:
1339 <ul>
1340 <li><font color="#3366ff">some fixes</font></li>
1341 <li><font color="#ff0000">some fixes for VBR</font></li>
1342 </ul>
1343 </li>
1344 <li>
1345 Gabriel Bouvigne:
1346 <ul>
1347 <li>--noasm switch. Might help Cyrix/Via users</li>
1348 <li><font color="#ff0000">presets and
1349 alt-presets merged</font>
1350 </li>
1351 </ul>
1352 </li>
1353 </ul>
1354 <br>
1355 <h3>LAME 3.92 &nbsp;April 14 2002</h3>
1356 <ul>
1357 <li><font color="#ff0000">Alexander
1358 Leidinger: add non linear psymodel (compile time option,
1359 disabled by default)</font>, workaround a bug in gcc 3.0.3
1360 (compiler options, based upon suggestions from various people, see
1361 archives and changelog for more)
1362 </li>
1363 <li>Steve Lhomme: ACM wrapper (MS-Windows codec)</li>
1364 <li><font color="#3366ff">Steve Lhomme:&nbsp;
1365 less memory copying on stereo (interleaved) input</font>
1366 </li>
1367 <li><font color="#ff0000">Takehiro Tominaga:
1368 Inter-channel masking, enables with --interch x option</font>
1369 </li>
1370 <li>For buggy versions of gcc compiler (2.96*), back off on
1371 some of the advanced compiler options<br>
1372 </li>
1373 </ul>
1374 <br>
1375 <h3>LAME 3.91 &nbsp;December 29 2001</h3>
1376 <ul>
1377 <li><font color="#ff0000">Darin
1378 Morrison: Bugfix for --alt-preset (for content with low
1379 volume, clean vocals), only important for the "fast standard" preset</font>
1380 </li>
1381 <li>
1382 Alexander Leidinger:
1383 <ul>
1384 <li>add some missing files to the distribution</li>
1385 <li>add --alt-preset to the man page</li>
1386 </ul>
1387 </li>
1388 </ul>
1389 <br>
1390 <h3>LAME 3.90 &nbsp;December 21 2001</h3>
1391 <ul>
1392 <li><font color="#ff0000">Many small improvements
1393 and bug fixes not added to history</font>
1394 </li>
1395 <li><font color="#ff0000">John
1396 Dahlstrom: more fine tuning on the auto adjustment of the ATH</font>
1397 </li>
1398 <li><font color="#3366ff">Robert
1399 Hegemann: small speed and quality improvements for the old
1400 VBR code (--vbr-old).</font>
1401 </li>
1402 <li><font color="#ff0000">Robert
1403 Hegemann: some short block bug fixes</font>
1404 </li>
1405 <li><font color="#ff0000">Robert
1406 Hegemann: Big improvements to --vbr-mtrh, now encodes much
1407 more frequencies over 16khz</font>
1408 </li>
1409 <li><font color="#ff0000">Robert
1410 Hegemann: --vbr-new code disabled (outdated and lower
1411 quality) and replaced with --vbr-mtrh (Both --vbr-new and --vbr-mtrh
1412 now default to mtrh)</font>
1413 </li>
1414 <li>Robert Hegemann: reordering of --longhelp to give
1415 more information, --extrahelp dropped
1416 </li>
1417 <li>
1418 Darin Morrison: Totally revamped and extremely
1419 high quality unified preset system and other general quality
1420 improvements now available with --alt-presets:
1421 <ul>
1422 <li>
1423 <font color="#ff0000">some improvements
1424 to psychoacoustics (vast improvements over default L.A.M.E. modes) when
1425 --alt-preset is used including:</font>
1426 <ul>
1427 <li><font color="#ff0000">Improved tuning
1428 of short block usage.</font>
1429 </li>
1430 <li><font color="#ff0000">Improved
1431 quantization selection usage (the -X modes), now adapts between
1432 appropriate modes on the fly. Also helps on "dropout" problems and with
1433 pre-echo cases.</font>
1434 </li>
1435 <li><font color="#ff0000">Improved joint
1436 stereo usage. Thresholds are better tuned now and fix some "dropout"
1437 problems L.A.M.E. suffers from on clips like serioustrouble.</font>
1438 </li>
1439 <li><font color="#ff0000">Improved noise
1440 shaping usage. Now switches between noise shaping modes on the fly
1441 (toggles -Z on and off when appropriate) which allows lower bitrates
1442 but without the quality compromise.</font>
1443 </li>
1444 <li><font color="#ff0000">Clips vastly
1445 improved over default L.A.M.E. modes (vbr/cbr/abr, including --r3mix):
1446 castanets, florida_seq, death2, fatboy, spahm, gbtinc, ravebase, short,
1447 florida_seq, hihat, bassdrum, 2nd_vent_clip, serioustrouble, bloodline,
1448 and others. No degraded clips known.</font>
1449 </li>
1450 <li>VBR bitrates are now more "stable" with less
1451 fluctuation -- not dipping too low on some music and not increasing too
1452 high unnecessarily on other music. "--alt-preset standard" provides
1453 bitrates roughly within the range of 180-220kbps, often averaging close
1454 to 192kbps.
1455 </li>
1456 </ul>
1457 </li>
1458 <li>--alt-presets replace the --dm-presets and "metal"
1459 preset is removed and replaced with generic abr and cbr presets.
1460 </li>
1461 <li>--alt-preset extreme (note the 'e') replaces xtreme to
1462 help eliminate some confusion
1463 </li>
1464 <li>--alt-preset vbr modes now have a fast option which
1465 offers almost no compromise in speed.
1466 </li>
1467 <li>--alt-preset standard (and "fast standard") are now
1468 much lower in bitrate, matching --r3mix with an overall average, though
1469 offering higher quality especially on difficult test samples.
1470 </li>
1471 <li>--alt-presets are no longer just "presets" as in a
1472 collection of switches, instead they are now quality "modes" because of
1473 special code level tunings (those mentioned above).
1474 </li>
1475 <li>Use --alt-preset help for more information.</li>
1476 </ul>
1477 </li>
1478 <li>Roel VdB: more tuning on the --r3mix preset </li>
1479 <li>Jon Dee, Roel VdB: INFO tag</li>
1480 <li>Alexander Leidinger, [email protected]: added
1481 --scale-l and --scale-r to scale stereo channels independantly
1482 </li>
1483 <li>Takehiro Tominaga: <font color="#ff0000">new
1484 noise shaping mode, offering more "cutting edge" shaping according to
1485 masking, enabled via -q0</font>
1486 </li>
1487 <li>Mark Taylor: More work on --nogap </li>
1488 <li>Gabriel Bouvigne: Small changes to abr code for
1489 more accurate final bitrate
1490 </li>
1491 <li>Gabriel Bouvigne, [email protected]:&nbsp;
1492 Preliminary <a href="http://www.replaygain.org">
1493 ReplayGain</a> analysis code added (not functional yet)
1494 </li>
1495 <li>Gabriel Bouvigne, Alexander Leidinger:&nbsp;
1496 Documentation updates
1497 </li>
1498 <li>John Dahlstrom, [email protected]: floating point
1499 interface function in the Windows DLL
1500 </li>
1501 </ul>
1502 <br>
1503 <h3>LAME 3.89beta &nbsp;July 5 2001</h3>
1504 <ul>
1505 <li>John Stewart: long filename support for Win9x/NT.</li>
1506 <li>Takehiro Tominaga: LAME can calculate the CRC of
1507 VBR header, so now "lame -pv" works fine.
1508 </li>
1509 <li><font color="#ff0000">Robert
1510 Hegemann: Improvements of the new VBR code (--vbr-mtrh).</font>
1511 </li>
1512 <li><font color="#3366ff">Robert Hegemann: New VBR
1513 code (--vbr-mtrh) is now defaulted to get more feedback. The VBR speed
1514 is now on par with CBR. We will use the old VBR code in the release.</font>
1515 </li>
1516 <li><font color="#ff0000">Gabriel Bouvigne: Change
1517 of the maximum frame size limit. LAME should now be more friendly with
1518 hardware players.</font>
1519 </li>
1520 <li>Gabriel Bouvigne: Size of VBR is now more balanced
1521 according to the -V value.
1522 </li>
1523 <li>Alexander Leidinger: Finished the implementation of the
1524 set/get functions.
1525 </li>
1526 <li>John Dahlstrom: LAME now handles 24bits input</li>
1527 <li>Mark Taylor: bugs in lame --decode causing truncation of
1528 mp3 file fixed
1529 </li>
1530 <li>Mark Taylor: preliminary --nogap support</li>
1531 <li>"Final" API completed: shared library safe! &nbsp;This
1532 API is frozen and should be backwords compatiable with future versions
1533 of libmp3lame.so, but we will continue to add new functionality.
1534 &nbsp;<br>
1535 </li>
1536 </ul>
1537 <h3> LAME 3.88beta&nbsp; March 25 2001</h3>
1538 <ul>
1539 <li><font color="#ff0000">A lot of work that was
1540 never added to the History!</font>
1541 </li>
1542 <li><font color="#ff0000">Frank Klemm and
1543 Gabriel Bouvigne: New ATH formula. Big improvement
1544 for high bitrate encodings.</font>
1545 </li>
1546 <li><font color="#ff0000">Takehiro Tominaga:
1547 Temporal masking</font>
1548 </li>
1549 <li><font color="#ff0000">Gabriel Bouvigne/Mark
1550 Taylor: auto adjustment of ATH</font>
1551 </li>
1552 <li><font color="#ff0000">Robert
1553 Hegemann: Better outer_loop stopping criterion.&nbsp;
1554 Enabled with -q2 or better.</font>
1555 </li>
1556 <li><font color="#ff0000">Robert Hegemann/Naoki
1557 Shibata: slow/carefull noise shaping.&nbsp;
1558 -q3..9: amplify all distorted bands. -q2: amplify
1559 distorted bands within 50%. -q1-0: amplify only
1560 most distorted band at each iteration.</font>
1561 </li>
1562 <li><font color="#ff0000">Takehiro Tominaga:
1563 Interframe, shortblock temporal masking.</font>
1564 </li>
1565 <li>Takehiro Tominaga: LAME restructured into a
1566 shared library and front end application. Slight changes to
1567 the API. More changes are coming to turn LAME into a true shared
1568 library (right now you have to recompile if you upgrade the library :-(
1569 </li>
1570 <li>
1571 <font color="#000000">Naoki Shibata:</font>
1572 <ul>
1573 <li><font color="#ff0000">improvements to
1574 psychoacoustics</font><font color="#000000">
1575 (--nspsytune)</font>
1576 <li><font color="#ff0000">BUG in long block
1577 pre echo control fixed </font><font color="#000000">
1578 (some out of range array access in M/S psychoacoustics)</font>
1579 </li>
1580 </ul>
1581 </li>
1582 <li><font color="#000000">Ralf
1583 Kempkens:&nbsp;&nbsp; Visual Basic Script for lame,
1584 suggested to put it on your Windows Desktop and you can drag'n'drop
1585 Waves to encode on it.</font>
1586 </li>
1587 <li><font color="#000000">Alexander
1588 Stumpf:&nbsp;&nbsp; improved lame.bat for 4Dos users</font>
1589 </li>
1590 <li><font color="#000000">Mark Taylor: Several
1591 bugs fixed in the resampling code.</font>
1592 </li>
1593 <li><font color="#000000">Frank Klemm, Robert
1594 Hegemann:&nbsp;&nbsp; added assembler code for CPU
1595 feature detection on runtime (MMX, 3DNow, SIMD)</font>
1596 </li>
1597 <li><font color="#3366ff">Takehiro Tominaga:
1598 3DNow FFT code.</font>
1599 </li>
1600 <li><font color="#000000">Florian Bome,
1601 Alexander Leidinger:&nbsp;&nbsp; more work on
1602 configure stuff</font>
1603 </li>
1604 <li><font color="#000000">Alexander
1605 Leidinger:&nbsp; automake/libtool generated Makefiles and
1606 TONS of other work.</font>
1607 </li>
1608 <li><font color="#000000">Alexander
1609 Leidinger:&nbsp; Much work towards shared library style
1610 API.</font>
1611 </li>
1612 <li><font color="#000000">Anonymous: New more
1613 efficient RTP code.</font>
1614 </li>
1615 <li><font color="#ff0000">Mark Taylor:
1616 psycho-acoustic data now computed for all scalefactor bands (up to 24
1617 kHz)</font>
1618 </li>
1619 <li><font color="#ff0000">Mark Taylor, Takehiro
1620 Tominaga: All ISO table data replaced by formulas - should improve
1621 MPEG2.5 results for which we never had correct table data.</font>
1622 </li>
1623 </ul>
1624 <h3> LAME 3.87alpha &nbsp;September 25 2000</h3>
1625 <ul>
1626 <li>Mark Taylor: Bug fixed in LAME/mpglib error
1627 recovery when encountering a corrupt MP3 frame during
1628 *decoding*.
1629 </li>
1630 <li>Albert Faber: added LayerI+II decoding support</li>
1631 <li><font color="#000000">Frank Klemm:&nbsp;
1632 added improved CRC calculation</font>
1633 </li>
1634 <li><font color="#000000">Frank Klemm:
1635 substantial code cleanup/improvements</font>
1636 </li>
1637 <li>
1638 Robert Hegemann: Bug fixes
1639 <ul>
1640 <li><font color="#ff0000">in huffman_init</font>,
1641 could lead to segmentation faults (only in rare cases, most likely at
1642 lower sample rates)
1643 </li>
1644 <li><font color="#ff0000">M/S switching at
1645 lower sample rates</font> (the fact there is no 2nd granule was
1646 ignored)
1647 </li>
1648 </ul>
1649 </li>
1650 <li><font color="#3366ff">Robert
1651 Hegemann: speed up in VBR</font>
1652 </li>
1653 <li>Jarmo Laakkonen: Amiga/GCC settings for
1654 Makefile.unix.
1655 </li>
1656 <li>Magnus Holmgren: README and Makefile for (free)
1657 Borland C++ compiler. Will also compile lame_enc.dll, but
1658 this is untested.
1659 </li>
1660 <li>Florian Bome:&nbsp; LAME finally has
1661 a ./configure script!!
1662 </li>
1663 </ul>
1664 <h3> LAME 3.86beta &nbsp;August 6 2000</h3>
1665 <ul>
1666 <li>Christopher Wise: A makefile for DJGPP, the DOS
1667 version of gcc. Now most windows users should be able to
1668 compile LAME with minimal effort.
1669 </li>
1670 <li><font color="#ff0000">Robert
1671 Hegemann: old VBR:&nbsp; fixed some bugs and
1672 Takehiro's scalefac_scale feature (not yet on by&nbsp;
1673 default.) older LAME versions did not allow to spent more
1674 than 2500 bits of 4095 possible bits to a granule per channel, now
1675 fixed.</font>
1676 </li>
1677 <li>Robert Hegemann: new VBR:&nbsp;&nbsp;
1678 analog silence treatment like in old VBR
1679 </li>
1680 <li>William Welch: Improved options for Linux/Alpha
1681 gcc and ccc compilers in Makefile.
1682 </li>
1683 <li>Mathew Hendry: setting appropriate CRC bit for
1684 additional Xing-VBR tagging frame
1685 </li>
1686 <li>Don Melton: added ID3 version 2 TAG support</li>
1687 <li><font color="#000000">John Dahlstrom: fixed
1688 bug allowing timing information (for status in command line encoder) to
1689 overflow.</font>
1690 </li>
1691 <li><font color="#000000">Tamito KAJIYAMA, Fixed
1692 several bugs in the LAME/Vorbis interface.</font>
1693 </li>
1694 <li><font color="#000000">Mark Taylor:&nbsp;
1695 lame --decode will recognize <a href="http://albumid.cjb.net">Album
1696 ID tags</a></font>
1697 </li>
1698 <li><font color="#ff0000">Naoki
1699 Shibata: Additive masking and other improvements to psycho
1700 acoustics. (not yet on by default)</font>
1701 </li>
1702 </ul>
1703 <h3> LAME 3.85beta&nbsp; July 3 2000</h3>
1704 <ul>
1705 <li><font color="#ff0000">Takehiro
1706 Tominaga: mid/side stereo demasking thresholds updated.</font>
1707 </li>
1708 <li>Takehiro Tominaga: New short block MDCT coefficient data
1709 structure. Should allow for future speed improvements.
1710 </li>
1711 <li>Robert Hegemann: fixed bug in old VBR routine,
1712 the --noath mode messed up the VBR routine resulting in very large files
1713 </li>
1714 <li>Robert Hegemann: found bugs in some sections when using 32
1715 bit floating point. Default is now back to 64bit floating
1716 point.
1717 </li>
1718 <li><font color="#ff0000">Takehiro
1719 Tominaga: Modified PE formula to use ATH.</font>
1720 </li>
1721 <li><font color="#000000">S.T.L.:&nbsp;
1722 README.DJGPP - instructions for compiling LAME with DJGPP, the dos
1723 version of gcc.</font>
1724 </li>
1725 </ul>
1726 <h3> LAME 3.84beta &nbsp;June 30 2000</h3>
1727 <ul>
1728 <li>Mark Weinstein: .wav file output (with --decode
1729 option) was writing the wrong filesize in the .wav file. Now
1730 fixed.
1731 </li>
1732 <li>Mark Taylor: (optional) Vorbis support, both
1733 encoding and decoding. LAME can now produce .ogg files, or
1734 even re-encode your entire .ogg collection into
1735 mp3.&nbsp; (Just kidding: it is always a bad idea to
1736 convert from one lossy format to another)
1737 </li>
1738 <li>?: Bug fixed causing VBR to crash under
1739 windows.&nbsp; (pretab[] array overflow)
1740 </li>
1741 <li>Sergey Sapelin: Another bug found in the mpg123 MPEG2
1742 tables. Now fixed for the mpg123 based decoder in LAME.
1743 </li>
1744 <li>Marco Remondini: VBR histogram works in
1745 win32. compile with -DBRHIST -DNOTERMCAP
1746 </li>
1747 <li><font color="#ff0000">Takehiro
1748 Tominaga: LAME CBR will now use scalefac_scale to expand the
1749 dynamic range of the scalefactors.</font>
1750 </li>
1751 <li><font color="#000000">Iwasa Kazmi:&nbsp;
1752 Library improvements: exit()'s, printf, fprintf's are being
1753 replaced by interceptable macros.</font>
1754 </li>
1755 </ul>
1756 <h3> LAME 3.83beta &nbsp;May 19 2000</h3>
1757 <ul>
1758 <li><font color="#ff0000">Mark Taylor:&nbsp;
1759 Bug in buffering routines: in some cases, could cause
1760 MDCT to read past end of buffer. Rare in MPEG2,
1761 even more rare for MPEG1, but potentially serious!</font>
1762 </li>
1763 <li>Mark Taylor: MDCT/polyphase filterbank was not
1764 being "primed" properly. Does not effect output unless you
1765 set the encoder delay lower than the default of 576 samples.
1766 </li>
1767 <li><font color="#ff0000">Mark Taylor:&nbsp;
1768 "vdbj" and "Caster" found several VBR bugs (now
1769 fixed):&nbsp; 1. Analog silence detection only
1770 checked frequencies up to 16 kHz. 2. VBR mode could
1771 still somehow avoid -F mode. 3. VBR mode would
1772 ignore noise above 16 kHz (scalefactor band 22), Now calc_noise1 will
1773 compute the noise in this band when in VBR mode. Not
1774 calculated in CBR mode since CBR algorithm has no way of
1775 using this information.</font>
1776 </li>
1777 <li>Mark Taylor: scalefactor band 22 info
1778 (masking(=ATH), noise and energy) now displayed in frame
1779 analyzer.
1780 </li>
1781 <li><font color="#ff0000">VBR code ATH tuning
1782 was disabled by accident in 3.81, now fixed.</font>
1783 </li>
1784 <li><font color="#000000">Mark Taylor:&nbsp;
1785 lame --decode will produce .wav files. (oops - size is off by
1786 a factor of 4)</font>
1787 </li>
1788 </ul>
1789 <h3> LAME 3.82beta &nbsp;May 11 2000</h3>
1790 <ul>
1791 <li>Robert Hegemann: Fixed bug in high bitrate joint
1792 stereo encodings.
1793 </li>
1794 <li><font color="#3366ff">Naoki
1795 Shibata: new long block MDCT routine</font>
1796 </li>
1797 </ul>
1798 <h3> LAME 3.81beta &nbsp;May 8 2000</h3>
1799 <ul>
1800 <li>all ISO code removed!</li>
1801 <li><font color="#3366ff">Takehiro Tominaga and
1802 Naoki Shibata: new window subband routines.</font>
1803 </li>
1804 <li><font color="#000000">Naoki
1805 Shibata: Bug fix in mpglib (decoding) lib: in some
1806 cases, MDCT coefficients from previous granule was incorrectly used for
1807 the next granule.</font>
1808 </li>
1809 <li><font color="#ff0000">ISO 7680 bit buffer
1810 limitation removed. It can be reactivated with
1811 "--strictly-enforce-ISO" Please report any trouble with high
1812 bitrates.</font>
1813 </li>
1814 </ul>
1815 <h3> LAME 3.80beta &nbsp;May 6 2000</h3>
1816 <ul>
1817 <li><font color="#ff0000">Takehiro
1818 Tominaga: more efficient and faster huffman encoding!</font>
1819 </li>
1820 <li><font color="#ff0000">Takehiro Tominaga and
1821 Mark Taylor: much improved short block compression!</font>
1822 </li>
1823 <li><font color="#000000">Tomasz Motylewski and
1824 Mark Taylor: MPEG2.5 now supported!</font>
1825 </li>
1826 <li><font color="#000000">Mark Taylor:
1827 incorporated Takehiro's bitstream.c! bitstream.c used by
1828 default, but old ISO bitstream code can also be used.</font>
1829 </li>
1830 <li><font color="#ff0000">Scott Manley
1831 and Mark Taylor: good resampling routine finaly in
1832 LAME. uses a 19 point FIR filter with Blackman
1833 window. Very slow for non integer resampling ratios.</font>
1834 </li>
1835 <li><font color="#000000">Iwasa Kazmi:&nbsp;
1836 fixed SIGBUS error: VBR and id3 tags were using data after it
1837 was free()'d.</font>
1838 </li>
1839 <li><font color="#ff0000">Robert
1840 Hegemann: Improved VBR tuning. #define
1841 RH_QUALITY_CONTROL and #RH_SIDE_VBR now the defaults.</font>
1842 </li>
1843 <li><font color="#000000">Robert
1844 Hegemann:&nbsp; LAME version string now added to
1845 ancillary data.</font>
1846 </li>
1847 <li>Kimmo Mustonen: VBR histogram support for Amiga.</li>
1848 <li>Casper Gripenberg: VBR stats (but not histogram)
1849 for DOS verson.
1850 </li>
1851 <li>Robert Hegemann: rare VBR overflow bug fixed.</li>
1852 <li>Zack: -F option strictly enforces the VBR min
1853 bitrate. Without -F, LAME will ignore the minimum bitrate
1854 when encoding analog silence.
1855 </li>
1856 <li>Shawn Riley: User can now specify a compression
1857 ratio (--comp &lt;arg&gt;) instead of a bit rate.&nbsp;
1858 Default settings based on a compression ratio of 11.0
1859 </li>
1860 <li>Mark Taylor: free format bitstreams can be
1861 created with --freeformat, and specify any integer bitrate from 8 to
1862 320kbs with -b.
1863 </li>
1864 <li>Mark Taylor: lame be used as a decoder (output raw pcm
1865 only): lame --decode input.mp3 output.pcm
1866 </li>
1867 </ul>
1868 <h3> LAME 3.70&nbsp; April 6 2000</h3>
1869 <ul>
1870 <li>"LAME 3.69beta" becomes LAME 3.70 "stable"</li>
1871 </ul>
1872 <h3> LAME 3.69beta&nbsp; April 6 2000</h3>
1873 <ul>
1874 <li>"spahm": default mode selection bug
1875 fixed. In some cases, lame was defaulting to regular stereo
1876 instead of jstereo when the user did not specify a mode.
1877 </li>
1878 </ul>
1879 <h3> LAME 3.68beta &nbsp;April 4 2000</h3>
1880 <ul>
1881 <li>Mark Taylor: mono encoding bug in DLL fixed.</li>
1882 <li>Ingo Saitz: bug in --cwlimit argument parsing fixed.</li>
1883 <li><font color="#ff0000">Scott Manly: bug in
1884 4-point resample code fixed.</font>
1885 </li>
1886 </ul>
1887 <h3> LAME 3.67beta &nbsp;March 27 2000</h3>
1888 <ul>
1889 <li><font color="#ff0000">Robert
1890 Hegemann: jstereo now enabled for MPEG2 encodings</font>
1891 </li>
1892 <li>Mark Taylor: old M/S stereo mode which used L/R maskings
1893 has been removed.
1894 </li>
1895 <li>Mark Taylor: Xing MPEG2 VBR headers now working.</li>
1896 <li><font color="#ff0000">Mark Taylor:&nbsp;
1897 When quantized coefficients are all 0 in a band, set scalefactors to 0
1898 also to save a few bits.</font>
1899 </li>
1900 <li><font color="#000000">Ingo Saitz:&nbsp;
1901 Problems with framesize calculation when using -f fast-math option
1902 fixed.</font>
1903 </li>
1904 </ul>
1905 <h3> LAME 3.66beta &nbsp;March 21 2000</h3>
1906 <ul>
1907 <li>Bug fixes in BladeEnc DLL, possible click in last mp3
1908 frame, VBR historgram display, byteswapping option, ASM quantize
1909 routines work for both float and double.
1910 </li>
1911 </ul>
1912 <h3> LAME 3.65beta&nbsp; March 17 2000</h3>
1913 <ul>
1914 <li>Enabled ASM version of quantize_xrpow() - accidently
1915 disabled in lame3.64.
1916 </li>
1917 </ul>
1918 <h3> LAME 3.64beta &nbsp;March 16 2000</h3>
1919 <ul>
1920 <li>Don Melton: id3v1.1 tags &amp; id3 bugfixes</li>
1921 <li><font color="#ff0000">Gabriel
1922 Bouvigne: L/R matching block type fix</font>
1923 </li>
1924 <li><font color="#ff0000">Bug fixed which was
1925 allowing quantized values to exceed the maximum when not using -h</font>
1926 </li>
1927 <li><font color="#3366ff">Mark Taylor: Fitlers
1928 based on polyphase filterbank. should be slightly better
1929 since the responce is independent of the blocktype, and they are
1930 slightly faster.</font>
1931 </li>
1932 <li>Mark Taylor: API: the API changed slightly - and
1933 this should be the final version. There is a new routine:
1934 lame_encode_buffer() which takes an arbritray sized input buffer,
1935 resamples &amp; filters if necessary, encodes, and returns the
1936 mp3buffer. There are also several new #defines, so it is
1937 possible to compile a simple encoding library with no decoding or file
1938 I/O or command line parsing. see the file API for details.
1939 </li>
1940 <li>Mark Taylor: MSVC stuff: lame.exe (with and
1941 without the frame analyzer) and the CDex lame_enc.dll
1942 should compile under MSVC. The MSVC5 project files may need
1943 some tweaking. In particular,
1944 you need to make sure LAMEPARSE, LAMESNDFILE and HAVEMPGLIB
1945 are defined. (and HAVEGTK for the GTK stuff).
1946 </li>
1947 </ul>
1948 <h3>LAME 3.63beta &nbsp;February 20 2000</h3>
1949 <ul>
1950 <li>Robert Hegemann: FPE with -h fixed?</li>
1951 <li>Mathey Hendry: FPE error catching for Cygwin,
1952 FPE fix for vbr mode and output to /dev/null
1953 </li>
1954 <li>Jeremy Hall: Fixed problems with input files
1955 where the number of samples is not known.
1956 </li>
1957 <li><font color="#3366ff">Mathew
1958 Hendry: ASM quantize_xrpow() for GNU i386</font>
1959 </li>
1960 <li><font color="#3366ff">Wilfried
1961 Behne quantize_xrpow ()for PowerPC and non-ASM</font>
1962 </li>
1963 <li><font color="#3366ff">Takehiro
1964 Tominaga: GOGO FFTs (not yet used?)</font>
1965 </li>
1966 </ul>
1967 <h3>LAME 3.62beta&nbsp; February 9 2000</h3>
1968 <ul>
1969 <li><font color="#000000">Iwasa Kazmi:&nbsp;
1970 frame analyzer short block display of single subblocks (press
1971 1,2 or 3)</font>
1972 </li>
1973 <li><font color="#000000">Ingo Saitz:&nbsp;
1974 --help option added, with output to stdout</font>
1975 </li>
1976 <li><font color="#ff0000">Alfred Weyers: short
1977 block AAC spreading function bug fixed</font>
1978 </li>
1979 <li><font color="#3366ff">Takehiro
1980 Tominaga: new scalefac data structure - improves performance!</font>
1981 </li>
1982 <li><font color="#ff0000">Lionel
1983 Bonnet: Bug fixed in MPEG2 scalefactor routine: scalefactors
1984 were being severly limited.</font>
1985 </li>
1986 <li><font color="#3366ff">Takehiro
1987 Tominaga: faster FFT routines from. These routines
1988 are also compatible with the GOGO routines, in case someone is
1989 interested in porting them back to LAME.</font>
1990 </li>
1991 <li><font color="#3366ff">Sigbj&oslash;rn
1992 Skj&aelig;ret, Takehiro Tominaga: faster pow() code.</font>
1993 </li>
1994 <li><font color="#ff0000">Joachim
1995 Kuebart: Found some unitialized variables that were effecting
1996 quality for encodings which did not use the -h option (now fixed).</font>
1997 </li>
1998 <li>Mark Taylor: More modularization work.&nbsp;&nbsp;
1999 It is now possible to use LAME as a library where you can set the
2000 encoding parameters directly and do your own file
2001 i/o.&nbsp; The calling program is now it's own mp3
2002 output. For an example of the LAME API, see main.c, or
2003 mp3rtp.c or mp3x.c. These can all be compiled as stand alone
2004 programs which link with libmp3lame.a.
2005 </li>
2006 <li>Felix vos Leitner: mp3rtp fixes.&nbsp;
2007 mp3rtp is a standalone program which will encode and stream with RTP.
2008 </li>
2009 <li>Robert Hegemann: Information written to stderr
2010 displaying exactly which type of lowpass filter (if any) is being used.
2011 </li>
2012 <li>Iwasa Kazmi: mpglib (the mpg123 decoder) scsfi
2013 decoding fixes.
2014 </li>
2015 <li>Takehiro Tominaga: More mpglib scsfi decoding
2016 fixes.
2017 </li>
2018 </ul>
2019 <h3>LAME 3.61beta&nbsp; January 14 2000</h3>
2020 <ul>
2021 <li><font color="#ff0000">Mark Taylor: Fixed bug
2022 with lowpass filters when using VBR with a 64kbs or lower min bitrate
2023 setting.</font>
2024 </li>
2025 <li><font color="#ff0000">Takehiro
2026 Tominaga: more efficient huffman encoding splitting.</font>
2027 </li>
2028 </ul>
2029 <h3> LAME 3.60beta&nbsp; January 9 2000</h3>
2030 <ul>
2031 <li>Mark Taylor: Distribution now comes with self
2032 test. Needs work to be automated, see 'make test' in Makefile.
2033 </li>
2034 <li><font color="#ff0000">Mark Taylor: AAC
2035 spreading function now the default</font>
2036 </li>
2037 <li>Gabriel Bouvigne: updated HTML docs</li>
2038 <li>Felix von Leitner: compute correct file length from Xing
2039 header (if present) when input file is a mp3 file
2040 </li>
2041 <li>Felix von Leitner: mp3rtp (standalone) program now
2042 included. Not yet tested. mp3rtp
2043 ip:port:ttl &lt;infile&gt;&nbsp;&nbsp;
2044 /dev/null will stream directly to ip:port using RTP.
2045 </li>
2046 </ul>
2047 <h3>LAME 3.59beta &nbsp;January 4 2000</h3>
2048 <ul>
2049 <li>Takehiro Tominaga: --noath option.&nbsp;
2050 Disables ATH maskings.
2051 </li>
2052 <li>Gabriel Bouvigne: updated HTML docs.</li>
2053 <li>Iwasa Kazmi: makefile fixes</li>
2054 <li>Mark Taylor: Fixed bug where first frame of data
2055 was always overwritten with 0's. Thanks to 'gol'
2056 </li>
2057 <li><font color="#ff0000">Mark Taylor:&nbsp;
2058 bug fixes in mid/side masking ratios (thanks to Menno Bakker)</font>
2059 </li>
2060 <li>Mark Taylor: replaced norm_l, norm_s table data
2061 with formulas.
2062 </li>
2063 </ul>
2064 <h3>LAME 3.58beta &nbsp;December 13 1999</h3>
2065 <ul>
2066 <li><font color="#ff0000">Segher
2067 Boessenkool: More accurate quantization procedure!&nbsp;
2068 Enabled with -h.</font>
2069 </li>
2070 <li><font color="#3366ff">Mathew Hendry, Acy
2071 Stapp and Takehiro Tominaga: ASM optimizations for quantize_xrpow and
2072 quantize_xrpow_ISO.</font>
2073 </li>
2074 <li>Chuck Zenkus: "encoder inside" logo on web page</li>
2075 <li>Mark Taylor:&nbsp; a couple people have
2076 asked for this.&nbsp; Allow LAME to overide
2077 VBR_min_bitrate if analog_silence detected.&nbsp;&nbsp;
2078 Analog_silence defined a la Robert:&nbsp;&nbsp;
2079 energy &lt; ATH.
2080 </li>
2081 <li>An Van Lam: Valid bitrates were being printed for layer 2,
2082 not layer 3!
2083 </li>
2084 <li>Ethan Yeo: Makefile.MSVC updated</li>
2085 <li>Mark Stephens: updated all MSVC project files</li>
2086 <li>Robert Hegemann: lowpass and highpass filters
2087 can be enabled with --lowpass, --highpass
2088 </li>
2089 <li><font color="#ff0000">Mark Taylor:&nbsp;
2090 MS switching is now smoother: ms_ratio average over 4 granules</font>
2091 </li>
2092 <li><font color="#ff0000">Takehiro
2093 Tominaga: Scalefactor pre-emphasis fixed (and now turned back
2094 on)</font>
2095 </li>
2096 <li><font color="#ff0000">Takehiro
2097 Tominaga: Bug in M/S maskings: switch to turn on
2098 stereo demasking code was buggy.</font>
2099 </li>
2100 </ul>
2101 <h3>LAME 3.57beta &nbsp;November 22 1999</h3>
2102 <ul>
2103 <li>Sigbj&oslash;rn Skj&aelig;ret, patch to allow
2104 encoding from 8bit input files when using LIBSNDFILE
2105 </li>
2106 <li>Mark Taylor: Automatic downsampling to nearest valid
2107 samplerate.
2108 </li>
2109 <li>Mark Taylor: Scalefactor bands demarked on MDCT plot in
2110 frameanalyzer
2111 </li>
2112 <li>Mark Taylor: Scalefactor preemphasis disabled for
2113 now.&nbsp; The algorithm was often doing more harm than
2114 good.
2115 </li>
2116 </ul>
2117 <h3>LAME 3.56beta &nbsp;November 19 1999</h3>
2118 <ul>
2119 <li>Kimmo Mustonen: portabilty code cleanup.</li>
2120 <li>Vladimir Marek: id3 genre patch.</li>
2121 <li>Conrad Sanderson: new applypatch script.</li>
2122 <li>Mark Taylor: Initial window type now "STOP_TYPE" to reduce
2123 initial attenuation. This is needed because the new encoder
2124 delay is so short. With a NORM_TYPE, the first 240 samples
2125 would be attenuated.
2126 </li>
2127 <li>Mark Taylor: Padding at end of file now adjusted
2128 (hopefully!) to produce as little padding as possible while still
2129 guarantee all input samples are encoded.
2130 </li>
2131 <li><font color="#ff0000">Takehiro
2132 Tominaga: Reduced shortblock extra bit allocation formulas by
2133 10% since new huffman coding is at least 10% more efficient.</font>
2134 </li>
2135 </ul>
2136 <h3>LAME 3.55beta &nbsp;November 11 1999</h3>
2137 <ul>
2138 <li>Albert Faber: updated BladeEnc.dll</li>
2139 <li>
2140 Mark Taylor: Simple lowpass filter added to linear
2141 downsampling routine.
2142 </li>
2143 <li>Nils Faerber: updated man page.</li>
2144 <li>
2145 Mark Taylor: All floating point variables are delcared
2146 FLOAT or FLOAT8. Change the definition of FLOAT8 in
2147 machine.h to run at 32bit preceision.
2148 </li>
2149 <li>
2150 Mark Taylor: Bug (introduced in 3.54beta) in
2151 stereo-&gt;mono downsampling fixed.
2152 </li>
2153 </ul>
2154 <h3>LAME 3.54beta &nbsp;November 8 1999</h3>
2155 <ul>
2156 <li>
2157 Mark Taylor: Encoder delay is now 48 samples. Can be adjusted to
2158 1160 to sync with FhG (see ENCDELAY in encoder.h) This is kind
2159 of amazing, since if Takehiro put his MDCT/filterbank routine in
2160 a decoder, we could have a total delay of only 96 samples.
2161 </li>
2162 <li>
2163 <font color="#ff0000">Mark Taylor: More inconstancies found and
2164 fixed in MPEG2 tables.</font>
2165 </li>
2166 <li>
2167 Mark Taylor: Resampling from an MP3 input file now works. But we
2168 still dont have a lowpass filter so dont expect good results.
2169 </li>
2170 </ul>
2171 <h3>LAME 3.53beta &nbsp;November 8 1999</h3>
2172 <ul>
2173 <li>
2174 <font color="#3366ff">Takehiro Tominaga: Fixed MPEG2 problem in
2175 new MDCT routines. Takehiro's combined filterbank/MDCT routine is
2176 now the default. Removes all buffering from psymodel.c and the
2177 filterbanks/MDCT routines.</font>
2178 </li>
2179 </ul>
2180 <h3>LAME 3.52beta &nbsp;November 8 1999</h3>
2181 <ul>
2182 <li>
2183 By permission of copyright holders of all GPL code in
2184 LAME, all GPL code is now released under a modified version
2185 of the LGPL (see the README file)
2186 </li>
2187 <li>
2188 By popular demand, all C++ comments changed to C style
2189 comments
2190 </li>
2191 <li>
2192 Mark Taylor: Linear resampling now works. Use
2193 --resample to set an output samplerate different from the input
2194 samplerate. (doesn't seem to work with mp3 input files, and
2195 there is no lowpass filter, so dont expect good results just yet)
2196 </li>
2197 <li>
2198 <font color="#3366ff">Takehiro Tominaga: Faster Huffman encoding
2199 routines</font>
2200 </li>
2201 </ul>
2202 <font color="#3366ff">The following changes are disabled
2203 because of MPEG2 problems. But to try them, set MDCTDELAY=48
2204 in encoder.h, instead of MDCTDELAY=528.:</font>
2205 <ul>
2206 <li>
2207 <font color="#3366ff">Takehiro Tominaga: New MDCT routines with
2208 shorter delay (48 samples instead of 528) and even faster than
2209 the old routines.</font>
2210 </li>
2211 <li>
2212 <font color="#3366ff">Takehiro Tominaga: Removed extra buffering
2213 in psymodel.c</font>
2214 </li>
2215 </ul>
2216 <h3>LAME 3.51 &nbsp;November 7 1999</h3>
2217 <ul>
2218 <li>
2219 Takehiro Tominaga: Bug in quantize.c absolute threshold of
2220 hearing calculation for non-44.1 kHz input files.
2221 </li>
2222 </ul>
2223 <h3>LAME 3.50 &nbsp;November 1 1999</h3>
2224 <ul>
2225 <li>LAME 3.37beta becomes official LAME 3.50 release</li>
2226 </ul>
2227 <h3>LAME 3.37beta &nbsp;November 1 1999</h3>
2228 <ul>
2229 <li>
2230 <font color="#ff0000">Lionel Bonnet: Found severe bug in MPEG2
2231 Short block SNR.</font>
2232 </li>
2233 <li>Sergey Sapelin: VBR Toc improvement.</li>
2234 <li>Sergey Dubov: fskip() routine</li>
2235 <li>
2236 Conrad Sanderson: replacement for filterbank.c. Not much faster
2237 but amazingly simpler.
2238 </li>
2239 </ul>
2240 <h3>LAME 3.36beta &nbsp;October 25 1999</h3>
2241 <ul>
2242 <li>Albert Faber: more MSVC and BladeDLL updates</li>
2243 <li>Kimmo Mustonen: Much code cleanup and Amiga updates</li>
2244 <li>Anton Oleynikov: Borland C updates</li>
2245 <li>
2246 Mark Taylor: More stdin fixes: For some reason, forward
2247 fseek()'s would fail when used on pipes even though it is okay
2248 with redirection from "&lt;". So I changed all the forward
2249 fseek()'s to use fread(). This should improve stdin support for
2250 wav/aiff files. If you know the input file is raw pcm, you can
2251 still use the '-r' option to avoid *all* seeking of any kind.
2252 </li>
2253 </ul>
2254 <h3>LAME 3.35beta &nbsp;October 21 1999</h3>
2255 <ul>
2256 <li>
2257 <font color="#ff0000">Leonid Kulakov: Serious bug in MPEG2
2258 scalefactor band tables fixed.</font>
2259 </li>
2260 <li>
2261 Portability patches from: Anton Oleynikov, Sigbj&oslash;rn
2262 Skj&aelig;ret, Mathew Hendry, Richard Gorton
2263 </li>
2264 <li>Alfred Weyers: compiler options, updated timestatus.</li>
2265 <li>Albert Faber: BladeDll and other updates (new machine.h).</li>
2266 <li>Monty: updated Makefile to fix gcc inline math bug.</li>
2267 </ul>
2268 <h3>LAME 3.34beta &nbsp;October 12 1999</h3>
2269 <ul>
2270 <li>
2271 <font color="#ff0000">Mark Taylor: Bug fixed: minimum bitrate in
2272 VBR mode could be ignored for a few frames.</font>
2273 </li>
2274 <li><font color="#ff0000">Mark Taylor: New (minor) VBR tunings.</font></li>
2275 <li>
2276 Tim Ruddick: New wav/aiff header parsing routines. Better parsing
2277 and fewer fseek()'s.
2278 </li>
2279 <li>Anton Oleynikov: patches to work with Borland C</li>
2280 <li>
2281 <font color="#ff0000">Gabriel Bouvigne: Experimental voice option
2282 enabled with --voice</font>
2283 </li>
2284 </ul>
2285 <h3>LAME 3.33beta &nbsp;October 11 1999</h3>
2286 <ul>
2287 <li>
2288 <font color="#ff0000">Robert Hegemann: RH VBR mode now the
2289 default and only VBR mode. The new code will always quantize to
2290 0 distortion and the quality is increased by reducing the
2291 masking from the psy-model. -X0 is still the default for now.</font>
2292 </li>
2293 <li><font color="#ff0000">Robert Hegemann: new -X5 mode</font></li>
2294 <li>Mathew Hendry: New timing code, removes the need for HAVETIMES</li>
2295 <li>
2296 <font color="#3366ff">Mathew
2297 Hendry: assembler quantize_xrpow for Windows</font>
2298 </li>
2299 <li>Iwasa Kazmi: stdin/stdout patch for Windows</li>
2300 <li>
2301 Mark Taylor: New option: "--athonly" will ignore the psy-model
2302 output and use only the absolute threshold of hearing for the
2303 masking.
2304 </li>
2305 </ul>
2306 <h3>LAME 3.32beta &nbsp;October 8 1999</h3>
2307 <ul>
2308 <li>
2309 <font color="#3366ff">Takehiro Tominaga: faster long block
2310 spreading function convolution for non 44.1 kHz sampling
2311 frequencies, and faster short block spreading function
2312 convolution for all sampling frequencies.</font>
2313 </li>
2314 <li>
2315 <font color="#ff0000">Takehiro Tominaga: Completly rewritten
2316 huffman table selection and count_bits(). More efficient table
2317 selection results in many more bits per frame.</font>
2318 </li>
2319 <li>
2320 <font color="#ff0000">Takehiro Tominaga: More efficient scalefac
2321 compress setting.</font>
2322 </li>
2323 <li><font color="#3366ff">Mike Cheng: new calc_noise2()</font></li>
2324 <li>Alfred Weyers: patch for timestatus() seconds rollover</li>
2325 </ul>
2326 <h3>LAME 3.31beta &nbsp;September 28 1999</h3>
2327 <ul>
2328 <li>
2329 Albert Faber: updated his BladeDLL code. This allows LAME to be
2330 compiled into a BladeEnc compatiable .dll.
2331 </li>
2332 <li>
2333 <font color="#3366ff">Mike Cheng: faster l3psycho_ener()
2334 routine.</font>
2335 </li>
2336 <li>Sigbj&oslash;rn Skj&aelig;ret: more code cleanup.</li>
2337 </ul>
2338 <h3>LAME 3.30beta &nbsp;September 27 1999</h3>
2339 <ul>
2340 <li>
2341 Conrad Sanderson: ID3 tag code added (type 'lame' for
2342 instructions)
2343 </li>
2344 <li>new mdct.c from Mike Cheng (no faster, but much cleaner code)</li>
2345 <li>
2346 Mathew Hendry: Microsoft nmake makefile and a couple other
2347 changes for MSVC
2348 </li>
2349 <li>
2350 More modulization work: One input sound file interface handles
2351 mp3's, uncompressed audio, with or without LIBSNDFILE. Fixes
2352 (hopefully) a bunch of file I/O bugs introduced in 3.29 (Mark Taylor)
2353 </li>
2354 <li>
2355 LAME will now print valid samplerate/bitrate combinations
2356 (Mark Taylor)
2357 </li>
2358 <li>stdin/stdout fix for OS/2 (Paul Hartman)</li>
2359 <li>
2360 For mp3 input files, total frames estimated based on
2361 filesize and first frame bitrate. (Mark Taylor)
2362 </li>
2363 <li>
2364 Updated all functions with new style prototypes.
2365 (Sigbj&oslash;rn Skj&aelig;ret)
2366 </li>
2367 </ul>
2368 <h3>LAME 3.29beta &nbsp;September 21 1999</h3>
2369 <ul>
2370 <li>
2371 <font color="#ff0000">Bug in bigv_bitcount fixed. Loop.c was
2372 overestimating the number of bits needed, resulting in wasted
2373 bits every frame. (Leonid A. Kulakov)</font>
2374 </li>
2375 <li>
2376 <font color="#ff0000">Bug in *_choose_table() fixed&nbsp; These
2377 routines would not sellect the optimal Huffman table in some
2378 cases.&nbsp;&nbsp; (Leonid A. Kulakov)</font>
2379 </li>
2380 <li><font color="#ff0000">Tuning of ATH normalization (macik)</font></li>
2381 <li>
2382 Removed unused variables and fixed function prototypes
2383 (Sigbj&oslash;rn Skj&aelig;ret)
2384 </li>
2385 <li>
2386 Sami Farin sent a .wav file that LAME built in support choked
2387 on. I added a slightly more sophisticated wav header parsing to
2388 handle this, but if you have trouble, use libsndfile.
2389 </li>
2390 <li>
2391 Resampling hooks and options added. Buffering
2392 and resampling routines need to be written.
2393 </li>
2394 <li>
2395 LAME will now take an mp3 file as input. When resampling code
2396 is working, LAME will be able to (for example) convert a high
2397 bitrate stereo mp3 to a low bitrate mono mp3 for streaming.
2398 </li>
2399 </ul>
2400 <h3>LAME 3.28beta &nbsp;September 15 1999</h3>
2401 <ul>
2402 <li>
2403 <font color="#ff0000">Serious bug fixed in
2404 high frequency MDCT coefficients. Huffman coding was
2405 reversing the order of the count1 block quadruples.
2406 (Leonid A. Kulakov)</font>
2407 </li>
2408 <li>
2409 nint() problems under Tru64 unix fixed and preprocessor
2410 variable HAVE_NINT removed. (Bob Bell)
2411 </li>
2412 <li>
2413 Compiler warning fixes and code cleanup (Sigbj&oslash;rn
2414 Skj&aelig;ret, Lionel Bonnet)
2415 </li>
2416 <li>
2417 USAGE file now includes suggestions for
2418 downsampling. For low bitrate encodings, proper downsampling
2419 can give dramatically better results. (John Hayward-Warburton)
2420 </li>
2421 </ul>
2422 <h3>LAME 3.27beta &nbsp;September 12 1999</h3>
2423 <ul>
2424 <li>
2425 Several bugs in encode.c and l3bitstream.c fixed by Lionel
2426 Bonnet.
2427 </li>
2428 <li>
2429 Bugs in new VBR (#define RH) formula for mono input file
2430 and mid/side encoding fixed.
2431 </li>
2432 </ul>
2433 <h3>LAME 3.26beta &nbsp;September 10 1999</h3>
2434 <ul>
2435 <li>
2436 The "-m m" option (mono .mp3 file) will automatically mix
2437 left and right channels if the input file is stereo. (Alfred
2438 Weyers)
2439 </li>
2440 <li>
2441 <font color="#ff0000">New quant_compare algorithm (method for
2442 deciding which of two quantizations is better) enabled with -X4
2443 (Greg Maxwell)</font>
2444 </li>
2445 <li>
2446 <font color="#ff0000">New mid/side VBR bit allocation formula.
2447 Mid channel bits are set by the quality requirements, and then
2448 the side channel uses a reduced number of bits (in a proportion
2449 coming from the fixed bitrate code). This might not be optimal,
2450 but it should be pretty good and no one knows what the optimal
2451 solution should be. (Greg Maxwell)</font>
2452 </li>
2453 <li>
2454 <font color="#ff0000">New VBR (#define RH) tunings based on
2455 detailed listening tests by Macik and Greg Maxwell.</font>
2456 </li>
2457 <li>
2458 Sigbj&oslash;rn Skj&aelig;ret fixed several
2459 compiler warnings (which turned out to be potential bugs)
2460 </li>
2461 <li>Takehiro Tominaga fixed a low bitrate bug in reduce_side()</li>
2462 <li>Alfred Weyers fixed some buffer overflows.</li>
2463 <li>
2464 <font color="#ff0000">New ATH (absolute threshold of hearing)
2465 formula replaces buggy ISO code, and adds analog silence
2466 treatment (removal of coefficients below below ATH). These are
2467 turned on by default but have not been fully tested. (Robert
2468 Hegemann)</font>
2469 </li>
2470 <li>
2471 <font color="#ff0000">Bug in short block
2472 spreading function fixed. (Robert Hegemann)</font>
2473 </li>
2474 </ul>
2475 <h3>LAME 3.25beta &nbsp;August 22 1999</h3>
2476 <ul>
2477 <li>
2478 Sigbj&oslash;rn Skj&aelig;ret fixed a zero byte malloc call.
2479 This bug was introduced in 3.24 and causes problems on non
2480 Linux systems.
2481 </li>
2482 <li>
2483 Bit allocation routines would sometimes allocate more than
2484 4095 bits to one channel of one granule. A couple of people
2485 reported problems that might be caused by this, especially at
2486 higher bitrates.
2487 </li>
2488 <li>
2489 Nils Faerber updated the man page and fixed many of the
2490 compiler warnings.
2491 </li>
2492 </ul>
2493 <h3>LAME 3.24beta &nbsp;August 15 1999</h3>
2494 <ul>
2495 <li>
2496 This release contains the following new code (for
2497 developers) which is disabled by default:
2498 </li>
2499 <li>
2500 Hegemann: Completely overhauled VBR code. Now computes exact
2501 number of bits required for the given qualty and then quantized
2502 with the appropriate bitrate.
2503 </li>
2504 <li>Several new quantization quality measures.</li>
2505 </ul>
2506 <h3>LAME 3.23beta &nbsp;August 8 1999</h3>
2507 <ul>
2508 <li>
2509 Very nice continuously updated VBR histogram display from
2510 Iwasa Kazmi. (disabled with --nohist).
2511 </li>
2512 <li>
2513 More modulerization work. The encoding engine can now be
2514 compiled into libmp3lame, but the interface is awkward.
2515 </li>
2516 <li>
2517 <font color="#ff0000">Bug fixed in FFT Hann window formula
2518 (Leonid A. Kulakov).</font>
2519 </li>
2520 <li>
2521 New LAME logo on the download page. Created by Chris Michalisles.
2522 </li>
2523 <li>
2524 <font color="#ff0000">Several VBR algorithm
2525 improvements from Robert Hegemann. New quantization noise
2526 metrics and VBR quality measure takes into account mid/side
2527 encoding. Should produce smaller files with the same quality,
2528 especially when using jstereo.</font>
2529 </li>
2530 </ul>
2531 <h3>LAME 3.22beta &nbsp;July 27 1999</h3>
2532 <ul>
2533 <li>
2534 Downsampling (stereo to mono) bug with MPEG2 fixed. (Mike
2535 Oliphant)
2536 </li>
2537 <li>
2538 Downsampling now merges L &amp; R channels - before it
2539 only took the L channel.
2540 </li>
2541 <li>
2542 More modularization and code cleanup from Albert Faber and
2543 myself.
2544 </li>
2545 <li>
2546 Input filesize limit removed for raw pcm input
2547 files. For other file types, LAME will still only read the
2548 first 2^32 samples, (27 hours of playing time at 44.1 kHz).
2549 </li>
2550 </ul>
2551 <h3>LAME 3.21beta &nbsp;July 26 1999</h3>
2552 <ul>
2553 <li>
2554 <font color="#ff0000">Correct Mid/Side masking thresholds for
2555 JSTEREO mode! This is enabled with-h. It makes LAME about 20%
2556 slower since it computes psycho-acoustics for L,R Mid and Side
2557 channels.</font>
2558 </li>
2559 <li>
2560 <font color="#ff0000">"Analog silence"
2561 threshold added. Keeps VBR from upping the bitrate during
2562 very quite passages. (Robert.Hegemann)</font>
2563 </li>
2564 <li>
2565 <font color="#ff0000">New VBR quality setting from Robert
2566 Hegemann. It is based on the idea that distortion at lower bit
2567 rates sounds worse than at higher bitrates, and so the allowed
2568 distortion (VBR quality setting) is proportional to the bitrate.
2569 Because of this, default minimum bitrate is now 32kbs.</font>
2570 </li>
2571 <li>
2572 <font color="#ff0000">Expermental subblock gain code enabled
2573 with -Z.</font>
2574 </li>
2575 <li>
2576 New "-r" option for raw pcm input files. With -r, LAME will not
2577 do any fseek()'s or look for wav and aiff headers on
2578 the input file.
2579 </li>
2580 <li>
2581 Bug fixes in mp3x (frame analyzer) for viewing frames near
2582 end of the file.
2583 </li>
2584 <li>
2585 Bug fixed to allow setting the sampling rate of raw pcm
2586 input files.
2587 </li>
2588 </ul>
2589 <h3>LAME 3.20beta &nbsp;July 19 1999</h3>
2590 <ul>
2591 <li>
2592 Bug in get_audio.c fixed. Libsndfile wrappers would not compile
2593 (Miguel Revilla Rodriguez)
2594 </li>
2595 <li>
2596 Nils Faerber found some unitialized variables and some
2597 wierd extranous computations in filter_subband, now fixed.
2598 This was causing seg faults on some machines.
2599 </li>
2600 </ul>
2601 <h3> LAME 3.19beta &nbsp;July 18 1999</h3>
2602 <ul>
2603 <li>
2604 <font color="#ff0000">Oops! Robert Hegemann immediatly found a
2605 bug in the new (old -Z option) quantization code. calc_noise1
2606 was not returning tot_noise, so non ms-stereo frames were
2607 buggy.</font>
2608 </li>
2609 </ul>
2610 <h3>LAME 3.18beta &nbsp;July 17 1999</h3>
2611 <ul>
2612 <li>
2613 <font color="#ff0000">Many psycho-acoustic bug fixes. Dan
2614 Nelson discovered a bug in MPEG2: For short blocks, the code
2615 assumes 42 partition bands. MPEG1 sometimes has less, MPEG2
2616 can have more. In MPEG1, this bug would not have effected the
2617 output if your compiler initializes static variables to 0 on
2618 creation. In MPEG2 it leads to array out-of-bounds access
2619 errors. Finally, there was a related bug in MPEG1/MPEG2, short
2620 &amp; long blocks where the energy above 16 kHz was all added to
2621 partition band 0. (the lowest frequeny partition band!)</font>
2622 </li>
2623 <li>
2624 <font color="#ff0000">The -Z option (Gabriel Bouvigne's idea of
2625 using total quantization noise to choose between two
2626 quantizations with the same value of "over") is now the default.
2627 I believe this helps remove the trilling sound in Jan's
2628 testsignal4.wav. The quality of testsignal2.wav and
2629 testsignal4.wav are now better than Xing and getting closer to FhG.</font>
2630 </li>
2631 <li>
2632 Bug fixes in frame &amp; sample count for downsampling
2633 mode. (ben "jacobs")
2634 </li>
2635 <li>Patches to improve modulization. (ben "jacobs")</li>
2636 </ul>
2637 <h3>LAME 3.17beta &nbsp;July 11 1999</h3>
2638 <ul>
2639 <li>
2640 substantial code cleanup towards goal of making LAME more
2641 modular.
2642 </li>
2643 </ul>
2644 <h3>LAME 3.16beta &nbsp;July 11 1999</h3>
2645 <ul>
2646 <li>
2647 <font color="#ff0000">New tunings of window switching, and
2648 better bit allocation based on pe. (Jan Rafaj. improves both
2649 testsignal2.wav and testsignal4.wav).</font>
2650 </li>
2651 <li>
2652 <font color="#ff0000">Bug in mid/side quantization when side
2653 channel was zero fixed. (Albert Faber)</font>
2654 </li>
2655 <li>
2656 Removed some extranous computations in l3psy.c (Robert
2657 Hegemann)
2658 </li>
2659 <li>
2660 More detailed timing status info, including hours display.
2661 (Sakari Ailus) and percentage indicator (Conrad Sanderson).
2662 </li>
2663 <li>
2664 <font color="#3366ff">Window_subband and calc_noise1,
2665 calc_noise2 speedups. Quantize_xrpow speedup should be
2666 significant on non GNU/intel systems. (Mike Cheng)</font>
2667 </li>
2668 <li>
2669 <font color="#3366ff">Better initial guess for VBR bitrate.
2670 Should speed up VBR encoding. (Gabriel Bouvigne)</font>
2671 </li>
2672 <li>
2673 More advanced .wav header parsing. fixes bugs involving click
2674 in first frame. (Robert.Hegemann)
2675 </li>
2676 <li>
2677 Correct filesize and total frame computation when using
2678 LIBSNDFILE (ben "jacobs")
2679 </li>
2680 <li>
2681 Click in last frame (buffering problem) when using
2682 libsndfile fixed.
2683 </li>
2684 <li>
2685 Audio I/O code overhauled. There is now a uniform audio i/o
2686 interface to libsndfile or the LAME built in wav/aiff
2687 routines. All audio i/o code localized to get_audio.c.
2688 </li>
2689 </ul>
2690 <h3>LAME 3.15beta</h3>
2691 <ul>
2692 <li>
2693 times()/clock() problem fixed for non-unix OS. (Ben "Jacobs")
2694 </li>
2695 <li>
2696 Fixed uninitialized pe[] when using fast mode. (Ben "Jacobs")
2697 </li>
2698 </ul>
2699 <h3>LAME 3.13 &nbsp;June 24 1999</h3>
2700 <ul>
2701 <li>Patches for BeOS from Gertjan van Ratingen.</li>
2702 <li>Makefile info for OS/2 Warp 4.0 (from dink.org).</li>
2703 <li>Status display now based on wall clock time, not cpu time.</li>
2704 <li>
2705 mem_alloc no longer allocates twice as much memory as
2706 needed (Jan Peman).
2707 </li>
2708 </ul>
2709 <h3>LAME 3.12pre9</h3>
2710 <ul>
2711 <li>
2712 Updated BLADEDLL code to handle recent changes (Albert
2713 Faber).
2714 </li>
2715 <li>
2716 Bug fixed in parsing options when not using GTK (Albert
2717 Faber).
2718 </li>
2719 <li>
2720 <font color="#ff0000">MPEG2 Layer III psycho acoustics now
2721 working.</font>
2722 </li>
2723 <li>
2724 <font color="#3366ff">Improved huffman encoding Chris
2725 Matrakidis. (10% faster). I dont know how he finds these
2726 improvements! LAME with full quality now encodes
2727 faster than real time on my PII 266.</font>
2728 </li>
2729 <li>
2730 Fixed time display when encoding takes more than 60
2731 minutes.
2732 </li>
2733 </ul>
2734 <h3>LAME 3.12pre8</h3>
2735 <ul>
2736 <li>
2737 <font color="#ff0000">New <a href="gpsycho/ms_stereo.html">
2738 mid/side stereo</a> criterion. LAME will use mid/side stereo
2739 only when the difference between L &amp; R masking thresholds
2740 (averaged over all scalefactors) is less then 5db. In several
2741 test samples it does a very good job mimicking the FhG
2742 encoder.</font>
2743 </li>
2744 <li>
2745 <font color="#ff0000">Bug in mid/side stereo fixed: independent
2746 variation of mid &amp; side channel scalefactors disabled.
2747 Because of the way outer_loop is currently coded, when encoding
2748 mid/side coefficietns using left/right thresholds, you have
2749 to vary the scalefactors simultaneously.</font>
2750 </li>
2751 <li>
2752 <font color="#ff0000">Bug in side/mid energy
2753 ratio calculation fixed. (Thanks to Robert Hegemann)</font>
2754 </li>
2755 <li>
2756 Default mode is stereo (not jstereo) if bitrate is chosen
2757 as 192kbs or higher. Tero Auvinen first pointed out that FhG
2758 seems to think at 160kbs, their encoder is so good it doesn't
2759 need jstereo tricks. Since LAME is not as good as FhG, I am
2760 going to claim that 192kbs LAME is so good it doens't need
2761 jstereo tricks, and thus it is disabled by default.
2762 </li>
2763 <li>
2764 WAV header parsing for big-endian machines, and automatic
2765 detection of big-endian machines. (Thanks to Sigbj&oslash;rn
2766 Skj&aelig;ret).
2767 </li>
2768 <li>added 56 sample delay to sync LAME with FhG.</li>
2769 <li>MP3x (frame analyzer) can now handle MPEG2 streams.</li>
2770 </ul>
2771 <h3>LAME 3.12pre7</h3>
2772 <ul>
2773 <li>
2774 MPEG2 layer III now working! lower bit rates (down to 8kbs)
2775 and 3 more sampling frequencies: 16000, 22050, 24000Hz. Quality
2776 is poor - the psy-model does not yet work with these
2777 sampling frequencies.
2778 </li>
2779 <li>
2780 Fixed "ERROR: outer_loop(): huff_bits &lt; 0." bug
2781 when using VBR.
2782 </li>
2783 <li>
2784 bash and sh scripts to run LAME on multiple files now
2785 included. (from Robert Hegemann and Gerhard Wesp respectively)
2786 </li>
2787 <li>
2788 bug fix in encoding times for longer files from&nbsp;
2789 (Alvaro Martinez Echevarria)
2790 </li>
2791 <li>yet another segfault in the frame analyzer fixed.</li>
2792 <li>
2793 ISO psy-model/bit allocation routines removed. This allowed
2794 makeframe() to be made much simpler, and most of the complicated
2795 buffering is now gone. Eventually I would like the encoding
2796 engine to be a stand alone library.
2797 </li>
2798 </ul>
2799 <h3>LAME 3.12pre6</h3>
2800 <ul>
2801 <li>
2802 <font color="#ff0000">Better VBR
2803 tuning. Find minimum bitrate with distortion less than the
2804 allows maximum. A minimum bit rate is imposed on frames with
2805 short blocks (where the measured distortion can not be
2806 trusted). A minimum frame bitrate can be specified
2807 with -b, default=64kbs.</font>
2808 </li>
2809 <li>
2810 <a href="http://www.zip.com.au/%7Eerikd/libsndfile">LIBSNDFILE</a>
2811 support. With libsndfile, LAME can encode almost all sound
2812 formats. Albert Faber did the work for this, including
2813 getting libsndfile running under win32.
2814 </li>
2815 <li>CRC checksum now working! (Thanks to Johannes Overmann)</li>
2816 <li>frame analyzer will now work with mono .mp3 files</li>
2817 <li><font color="#3366ff">more code tweeks from Jan Peman.</font></li>
2818 <li>
2819 <font color="#3366ff">Compaq-Alpha(Linux) fixes and speedups
2820 from Nils Faerber.</font>
2821 </li>
2822 <li>
2823 <font color="#3366ff">Faster bin_search_StepSize from Juha
2824 Laukala.</font>
2825 </li>
2826 <li>
2827 <font color="#3366ff">Faster quantize() from Mike Cheng</font>
2828 </li>
2829 <li>
2830 <font color="#3366ff">Faster
2831 quantize_xrpow() from Chris Matrakidis. xrpow_flag removed
2832 since this option is now on by default.</font>
2833 </li>
2834 <li>Fixed .wav header parsing from Nils Faerber.</li>
2835 <li>
2836 Xing VBR frame info header code from Albert
2837 Faber. "Xing" and "LAME 3.12" embedded in first
2838 frame.
2839 </li>
2840 <li>
2841 <font color="#ff0000">Bug in VBR bit allocation based on "over"
2842 value fixed.</font>
2843 </li>
2844 </ul>
2845 <h3>LAME 3.11 &nbsp;June 3 1999</h3>
2846 <ul>
2847 <li>Almost all warnings (-Wall) now fixed! (Thanks to Jan Peman)</li>
2848 <li>
2849 More coding improvements from Gabriel Bouvigne and Warren
2850 Toomey.
2851 </li>
2852 <li>
2853 <font color="#ff0000">VBR (variable bit rate). Increases bit
2854 rate for short blocks and for frames where the number of bands
2855 containing audible distortion is greater than a given value.
2856 Much tuning needs to be done.</font>
2857 </li>
2858 <li>Patch to remove all atan() calls from James Droppo.</li>
2859 </ul>
2860 <h3>LAME 3.10 &nbsp;May 30 1999</h3>
2861 <ul>
2862 <li>
2863 <font color="#3366ff">Fast mode
2864 (-f) disables psycho-acoustic model for real time encoding on
2865 older machines. Thanks to Lauri Ahonen who first sent a patch
2866 for this.</font>
2867 </li>
2868 <li>
2869 <font color="#ff0000">New bit reservoir usage scheme to
2870 accommodate the new pre-echo detection formulas.</font>
2871 </li>
2872 <li>
2873 <font color="#ff0000">Tuning of AWS and ENER_AWS pre-echo
2874 formulas by Gabriel Bouvigne and myself. They work great! now
2875 on by default.</font>
2876 </li>
2877 <li>
2878 In jstereo, force blocktypes for left &amp; right
2879 channels to be identical. FhG seems to do this. It
2880 can be disabled with "-d".
2881 </li>
2882 <li>Patches to compile MP3x under win32 (Thanks to Albert Faber).</li>
2883 <li>
2884 <font color="#3366ff">bin_serach_stepsize
2885 limited to a quantizationStepSize of -210 through 45.</font>
2886 </li>
2887 <li>
2888 <font color="#ff0000">outer_loop() will now vary Mid &amp; Side
2889 scalefactors independently. Can lead to better quantizations,
2890 but it is slower (twice as many quantizations to look at).
2891 Running with "-m f" does not need this and will run at the
2892 old speed</font>
2893 </li>
2894 <li>
2895 <font color="#ff0000">Bug in inner_loop
2896 would allow quantizations larger than allowed. (introduced in
2897 lame3.04, now fixed.)</font>
2898 </li>
2899 <li>Updated HTML documentation from Gabriel Bouvigne.</li>
2900 <li>Unix man page from William Schelter.</li>
2901 <li>
2902 <font color="#ff0000">numlines[] bug fixed. (Thanks to Rafael
2903 Luebbert, MPecker author).</font>
2904 </li>
2905 <li>
2906 <font color="#3366ff">Quantization speed improvements from Chris
2907 Matrakidis.</font>
2908 </li>
2909 <li>
2910 <font color="#ff0000">When comparing quantizations with the same
2911 number of bands with audible distortion, use the one with the
2912 largest scalefactors, not the first one outer_loop
2913 happened to find.</font>
2914 </li>
2915 <li>
2916 Improved defination of best quantization when using -f
2917 (fast mode).
2918 </li>
2919 <li>
2920 subblock code now working. But no algorithm to
2921 choose subblock gains yet.
2922 </li>
2923 <li>
2924 Linux now segfaults on floating point exceptions. Should
2925 prevent me from releasing binaries that crash on other
2926 operating systems.
2927 </li>
2928 </ul>
2929 <h3>LAME 3.04 &nbsp;May 22 1999</h3>
2930 <ul>
2931 <li>Preliminary documentation from Gabriel Bouvigne.</li>
2932 <li>
2933 <font color="#3366ff">I wouldn't have thought it was possible,
2934 but now there are even more speed improvements from Chris
2935 Matrakidis! Removed one FFT when using joint stereo, and many
2936 improvements in loop.c.</font>
2937 </li>
2938 <li>
2939 "Fake" ms_stereo mode renamed "Force" ms_stereo since it forces
2940 mid/side stereo on all frames. For some music this is said to be
2941 a problem, but for most music mode is probably better than the
2942 default jstereo because it uses specialized mid/side channel
2943 masking thresholds.
2944 </li>
2945 <li>Small bugs in Force ms_stereo mode fixed.</li>
2946 <li>Compaq Alpha fixes from Nathan Slingerland.</li>
2947 <li>
2948 <font color="#ff0000">Some new experimental pre-echo detection
2949 formulas in l3psy.c (#ifdef AWS and #ifdef ENER_AWS,
2950 both off by default. Thanks to Gabriel Bouvigne and Andre
2951 Osterhues)</font>
2952 </li>
2953 <li>
2954 Several bugs in the syncing of data displayed by mp3x (the
2955 frame analyzer) were fixed.
2956 </li>
2957 <li>
2958 highq (-h) option added. This turns on things
2959 (just one so far) that should sound better but slow down LAME.
2960 </li>
2961 </ul>
2962 <h3>LAME 3.03 &nbsp;May 18 1999 </h3>
2963 <ul>
2964 <li>
2965 <font color="#3366ff">Faster (20%) &amp;
2966 cleaner FFT (Thanks to Chris Matrakidis
2967 http://www.geocities.com/ResearchTriangle/8869/fft_summary.html)</font>
2968 </li>
2969 <li>
2970 mods so it works with VC++ (Thanks to Gabriel Bouvigne,
2971 www.mp3tech.org)
2972 </li>
2973 <li>
2974 MP3s marked "original" by default (Thanks to
2975 Gabriel Bouvigne, www.mp3tech.org)
2976 </li>
2977 <li>
2978 Can now be compiled into a BladeEnc compatible .DLL (Thanks to
2979 Albert Faber, CDex author)
2980 </li>
2981 <li>
2982 Patches for "silent mode" and stdin/stdout (Thanks to Lars Magne
2983 Ingebrigtsen)
2984 </li>
2985 <li>
2986 <font color="#ff0000">Fixed rare bug: if a long_block is
2987 sandwiched between two short_blocks, it must be changed to a
2988 short_block, but the short_block ratios have not been computed in
2989 l3psy.c. Now always compute short_block ratios just in case.</font>
2990 </li>
2991 <li>
2992 <font color="#ff0000">Fixed bug with initial quantize step size
2993 when many coefficients are zero. (Thanks to Martin Weghofer).</font>
2994 </li>
2995 <li>Bug fixed in MP3x display of audible distortion.</li>
2996 <li>improved status display (Thanks to Lauri Ahonen).</li>
2997 </ul>
2998 <h3>LAME 3.02 &nbsp;May 12 1999</h3>
2999 <ul>
3000 <li>
3001 <font color="#ff0000">encoder could use ms_stereo even if
3002 channel 0 and 1 block types were different. (Thanks to Jan
3003 Rafaj)</font>
3004 </li>
3005 <li>
3006 <font color="#ff0000">added -k option to
3007 disable the 16 kHz cutoff at 128kbs. This cutoff is never
3008 used at higher bitrates. (Thanks to Jan Rafaj)</font>
3009 </li>
3010 <li>
3011 <font color="#ff0000">modified pe bit allocation formula to
3012 make sense at bit rates other than 128kbs.</font>
3013 </li>
3014 <li>
3015 fixed l3_xmin initialization problem which showed up under
3016 FreeBSD. (Thanks to Warren Toomey)
3017 </li>
3018 </ul>
3019 <h3><b>LAME 3.01 &nbsp;May 11 1999</b> </h3>
3020 <ul>
3021 <li>max_name_size increased to 300 (Thanks to Mike Oliphant)</li>
3022 <li>patch to allow seeks on input file (Thanks to Scott Manley)</li>
3023 <li>
3024 fixes for mono modes (Thanks to everyone who pointed this out)
3025 </li>
3026 <li>overflow in calc_noise2 fixed</li>
3027 <li>
3028 it reservoir overflow when encoding lots of frames with
3029 all zeros (Thanks to Jani Frilander)
3030 </li>
3031 </ul>
3032 <hr>
3033 <h3>LAME 3.0 &nbsp;May 10 1999</h3>
3034 <ul>
3035 <li>
3036 <font color="#ff0000">added GPSYCHO (developed by Mark Taylor)
3037 </font>
3038 </li>
3039 <li>
3040 <font color="#000000">added MP3x (developed by Mark Taylor)</font>
3041 </li>
3042 <li>LAME now maintained by Mark Taylor</li>
3043 </ul>
3044 <h3>November 8 1998</h3>
3045 <ul>
3046 <li>Version 2.1f released</li>
3047 <li>
3048 50% faster filter_subband() routine in encode.c
3049 contributed by James Droppo
3050 </li>
3051 </ul>
3052 <h3>November 2 1998</h3>
3053 <ul>
3054 <li>Version 2.1e released.</li>
3055 <li>
3056 New command line switch <b>-a</b> auto-resamples a stereo input
3057 file to mono.
3058 </li>
3059 <li>
3060 New command line switch <b>-r</b> resamples from 44.1 kHz to 32
3061 kHz [this switch doesn't work really well. Very tinny sounding
3062 output files. Has to do with the way I do the resampling
3063 probably]
3064 </li>
3065 <li>
3066 Both of these were put into the ISO code in the encode.c file,
3067 and are simply different ways of filling the input buffers from
3068 a file.
3069 </li>
3070 </ul>
3071 <h3>October 31 1998</h3>
3072 <ul>
3073 <li>Version 2.1d released</li>
3074 <li>Fixed memory alloc in musicin.c (for l3_sb_sample)</li>
3075 <li>
3076 Added new command line switch (-x) to force swapping of
3077 byte order
3078 </li>
3079 <li>
3080 Cleaned up memory routines in l3psy.c. All the mem_alloc() and
3081 free() routines where changed so that it was only done <i>once</i>
3082 and not every single time the routine was called.
3083 </li>
3084 <li>
3085 Added a compile time switch -DTIMER that includes all timing
3086 info. It's a switch for the time being until some other people
3087 have tested on their system. Timing code has a tendency to do
3088 different things on different platforms.
3089 </li>
3090 </ul>
3091 <h3>October 18 1998</h3>
3092 <ul>
3093 <li>Version 2.1b released.</li>
3094 <li>Fixed up bug: all PCM files were being read as WAV.</li>
3095 <li>
3096 Played with the mem_alloc routine to fix crash under amigaos
3097 (just allocating twice as much memory as needed). Might see if
3098 we can totally do without this routine. Individual malloc()s
3099 where they are needed instead
3100 </li>
3101 <li>
3102 Put Jan Peman's quality switch back in. This reduces quality
3103 via the '-q&nbsp;&lt;int&gt;' switch. Fun speedup which is
3104 mostly harmless if you're not concerned with quality.
3105 </li>
3106 <li>Compiling with amiga-gcc works fine</li>
3107 </ul>
3108 <h3>October 16 1998</h3>
3109 <ul>
3110 <li>
3111 Version 2.1a released. User input/output has been cleaned up a
3112 bit. WAV file reading is there in a very rudimentary sense ie
3113 the program will recognize the header and skip it, but not read
3114 it. The WAV file is assumed to be 16bit stereo 44.1 kHz.
3115 </li>
3116 </ul>
3117 <h3>October 6 1998</h3>
3118 <ul>
3119 <li>
3120 Version 2.1 released with all tables now incorporated into
3121 the exe. Thanks to <b>Lars Magne Ingebrigtseni</b>
3122 </li>
3123 </ul>
3124 <h3>October 4 1998</h3>
3125 <ul>
3126 <li>
3127 In response to some concerns about the quality of the encoder, I
3128 have rebuilt the encoder from scratch and carefully compared output
3129 at all stages with the output of the unmodified ISO encoder.
3130 </li>
3131 <li>
3132 <a href="http://www.uq.net.au/%7Ezzmcheng/lame/download.html">
3133 Version 2.0</a> of LAME is built from the ISO source code (dist10),
3134 and incorporates modifications from myself and the 8hz effort. The
3135 output file from LAME v2.0 is <em>identical</em> to the output of
3136 the ISO encoder for my test file. Since I do not have heaps of time,
3137 I left the ISO AIFF file reader in the code, and did not incorporate
3138 a WAV file reader.
3139 </li>
3140 <li>
3141 Added section
3142 on <a href="http://www.uq.net.au/%7Ezzmcheng/lame/quality.html">
3143 quality</a>.
3144 </li>
3145 </ul>
3146 <h3> October 1 1998</h3>
3147 <ul>
3148 <li>Updated web page and released LAME v1.0</li>
3149 </ul>
3150 <hr>
3151 <h3>Up to September 1998</h3>
3152 <p>Working on the 8hz source code...</p>
3153 <ul>
3154 <li>
3155 Patched the <a href="http://www.8hz.com/">8hz</a> source
3156 code
3157 </li>
3158 <li>
3159 45% faster than original source (on my freebsd p166).
3160 <ul>
3161 <li>
3162 m1 - sped up the mdct.c and quantize() functions [MDCTD,
3163 MDCTD2, LOOPD]
3164 </li>
3165 <li>
3166 m2 - sped up the filter_subband routine using <b>Stephane
3167 Tavenard</b>'s work from musicin [FILTST]
3168 </li>
3169 <li>m2 - minor cleanup of window_subband [WINDST2]</li>
3170 <li>
3171 m2 - Cleaned up a few bits in l3psy.c. Replaced a sparse
3172 matrix multiply with a hand configured unrolling [PSYD]
3173 </li>
3174 <li>
3175 m3 - (amiga only) Added in the asm FFT for m68k (based on
3176 sources from <b>Henryk Richter</b> and <b>Stephane Tavenard</b>)
3177 </li>
3178 <li>m4 - raw pcm support back in</li>
3179 <li>
3180 m5 - put in a byte-ordering switch for raw PCM reading
3181 (just in case)
3182 </li>
3183 <li>
3184 m6 - reworked the whole fft.c file. fft now 10-15%
3185 faster.
3186 </li>
3187 <li>
3188 m7 - totally new fft routine. exploits fact that this is
3189 a real-&gt;complex fft. About twice as fast as previous
3190 fastest fft (in m6). (C fft routine is faster than the
3191 asm one on an m68k!)
3192 </li>
3193 <li>
3194 m8
3195 <ul>
3196 <li>
3197 Now encodes from stdin. Use '-' as the input
3198 filename. Thanks to <b>Brad Threatt</b>
3199 </li>
3200 <li>
3201 Worked out that the 1024point FFT only ever uses
3202 the first 6 phi values, and the first 465 energy
3203 values. Saves a bunch of calculations.
3204 </li>
3205 <li>
3206 Added a speed-up/quality switch. Speed is
3207 increased but quality is decreased <i>slightly</i>.
3208 My ears are bad enough not to be able to notice
3209 the difference in quality at low settings :).
3210 Setting '-q 1' improves speed by about 10%. '-q
3211 100' improves speed by about 26%. Enoding of my
3212 test track goes from 111s (at default '-q 0')
3213 to 82s (at -q 100). Thanks to <b>Jan Peman</b>
3214 for this tip.
3215 </li>
3216 </ul>
3217 </li>
3218 <li>
3219 m9 - fixed an error in l3psy.c. numlines[] is overwritten
3220 with incorrect data. Added a new variable numlines_s[]
3221 to fix this. Thanks again to <b>Jan Peman</b>.
3222 </li>
3223 <li>
3224 m10 - Down to 106 seconds by selecting a few more compiler
3225 options. Also added a pow20() function in l3loop.c to
3226 speed up (ever so slightly) calls to pow(2.0, x)
3227 </li>
3228 <li>
3229 m11
3230 <ul>
3231 <li>
3232 No speedups. Just cleaned up some bits of the code.
3233 </li>
3234 <li>
3235 Changed K&amp;R prototyping to 'normal' format.
3236 Thanks to <b>Steffan Haeuser</b> for his help here.
3237 </li>
3238 <li>
3239 Changed some C++ style comments to normal C comments in
3240 huffman.c
3241 </li>
3242 <li>
3243 Removed the #warning from psy_data.h (it was getting
3244 annoying!)
3245 </li>
3246 <li>
3247 Removed reference in bitstream.c to malloc.h. Is there a
3248 system left where malloc.h hasn't been superceded by
3249 stdlib.h?
3250 </li>
3251 </ul>
3252 </li>
3253 </ul>
3254 <li>
3255 In Progess:
3256 <ul>
3257 <li>
3258 my PSYD hack for the spreading functions is only valid for
3259 44.1 kHz - Should really put in a "if freq = 44.1 kHz"
3260 switch for it. Someone might want to extend the speedup for
3261 48 and 32 kHz.
3262 </li>
3263 <li>
3264 Putting in Jan Peman's quantanf_init speedup.
3265 </li>
3266 </ul>
3267 </li>
3268 </ul>
3269 <hr>
3270 <center>
3271 <p>
3272 <a href="http://validator.w3.org/check?uri=referer">
3273 <img src="images/valid-html401.png" alt="Valid HTML 4.01
3274 Transitional" height="31" width="88">
3275 </a>
3276 </p>
3277 </center>
3278 </body>
3279 </html>