MPEG Audio Decoder (MAD) v 0.15.0b has been released

http://www.underbit.com/products/mad/

"This release is a bit different from previous releases: there are now independent packages for libmad, libid3tag, and madplay.

There is not at this time an update to the MAD plug-in for Winamp; 0.14.2b is still the current release for the plug-in."

Summary of changelog in the next post

Here is a summary of the important changes:

General
- Shared libraries are now built by default.
- Cygwin builds have reverted to Cygwin dependency by default; MinGW
builds are still possible with the --host=mingw32 option to 'configure'.

libmad
- Minor compatibility enhancements: 64th entry to Layer I/Layer II
scalefactor table; allow reserved emphasis value in frame headers.
- Bug fix for ARM mad_f_scale64() that was tickled by certain optimization
options.
- API additions: mad_f_div(), MAD_EMPHASIS_RESERVED.
- Rewritten PPC fixed-point assembly.

libid3tag
- Numerous API additions: id3_tag_version(), id3_tag_options(),
id3_tag_setlength(), id3_frame_field(), id3_field_getlatin1(),
id3_field_getfulllatin1(), id3_genre_index(), id3_genre_number(),
id3_latin1_ucs4duplicate(), id3_utf16_ucs4duplicate(),
id3_utf8_ucs4duplicate().
- Fixed missing interfaces for other API routines.
- Bug fix for segfault rendering ID3v1 tags.
- Bug fix for reading ID3v2.2 frames.
- Other miscellaneous improvements.

madplay
- Support for arbitrary resampling ratios (--sample-rate).
- Support for RVA2 (relative volume adjustment) ID3 tag frames.
- Keyboard controls for dynamic volume control (+/-).
- Support for ancillary data output (--ancillary-output).
- New audio_qnx, audio_alsa modules (untested).
- New audio_carbon module (tested, for Mac OS X).
- Other miscellaneous improvements.

For complete details, see the CHANGES file in each package. Be sure
also to review the README in each package for modified build/install
instructions.

Nice. Can't wait for the Winamp 2.x plugin for this

Ciao..

any win. binarys around

OMG! n68 changed his posting style!

Hell freezes over.

Nice change, dude.

@rukes: I don't think the changes in libmad will make much difference for the winamp plugin. Nothing related to quality, at least.

how good is it compared to the tuned mpeglib123 decoder used in foobar?

It's worse. The dithering is lower quality and it can't noiseshape.

Can't believe! MAD is still alive!

Here's a Win32 binary. It's compiled with MSVC, but for some reason it doesn't take wildcards. There were no errors (nor warnings) when building it, so it should be OK.

All I really care about is that it can read from ID3v2 tags also. All my MP3's have ID3v2 tags, so the current MAD Winamp plugin cannot read them.

Ciao..


TnX.
@ speek.. is it tested..
and since there is quite a changelog.
is it running fine with your frontend..
(changes/features..)

Not really...

Ciao..


/edit/
i see you have it bundeled...
and the changes made are not to serious..
(frontend.)

I've just uploaded a build of 0.15.0b to Rarewares. Although it still uses the 0.14.2b code, I also included a build of the current Winamp2.x plugin recompiled against the 0.15.0 libs.

APE2 tag support would be cool if any1 could add it...

Ciao..


now there is 3. diff. mad builds for win..
and so far.. john`s compile is the fastest..

likely a dumb. question.:
i need mad to write a log.
how..? what`s the line..

I've made some mingw32 compiles, available here.

Hmm, yes my compile seems very slow. Better take one of the others. Good thing about Thop's mingw version is that it takes wildcards.

Just redirect stdout to a file:

Hmm i made some benchmarks with the different versions on a 96MB MP3 file using my AthlonXP 2224Mhz. Please see the quotes at the bottom to understand the difference between the mingw compiles. Here are the results:

MSVC
speek: 158s
john33: 80s

mingw32
gcc-2.95.3 default: 69s
gcc-2.95.3 speed: 55s
gcc-2.95.3 accuracy: 62s

gcc-3.2.3 default: 72s
gcc-3.2.3 speed: 57s
gcc-3.2.3 accuracy: 67s

It strikes me as odd that accuracy is faster than default, however i double-checked to make sure i didn't make a mistake during compiling or testing. madplay -V displays the optimization.

John33 isn't using MSVC, he's using Intel (ICL)

thanks, corrected. but notice rarewares says MSVC6 compile ...
corrected again

Ciao..


TnX.. dev0.

sometimes you don`t see
wood.. for all the tree's..

hum where is this release donwload page ?, on sourceforge there is no new package. maybe it's in the cvs ..

It's here:

ftp://ftp.mars.org/pub/mpeg/

Actually, he is!! That's an optimised MSVC6 compile. I use VC6 for MAD because I've had problems with using ICL in the past.

I don't understand why my GCC compile is so big compared to Thop's...

My madplay.exe is 487KB after packed with UPX...

I used GCC 3.2.3 with MSys 1.08

Can someone explain a little? Thanks a lot.

Edit: Figured it out...It is because I am a stupid...

what was it? forgot --disable-debugging?

Exactly.

I figured it out just after posting my first message.

Hi, john33:

Can you send your MSVC++ project file to me? It's very useful for me to create my QCD input plugin(the orig author didn't supply MSVC++ project).

You can send it here:
jin_xin_@hotmail.com

Rather than mail it to you, you, and anyone else who wants it, can download it here: http://homepage.ntlworld.com/jfe1205/MADprj.zip

THX:)
It's very very useful. Now I can throw the Cygwin away.

OMG, I compiled it with Visual Stdio .Net and got a lot of errors and warnings.
The worest is that I can't fix the error:(

How did you compile it with MSVC?

Maybe...Maybe...Maybe...I need your whole MAD project:)

john33, Can you give me a hand?

Don't panic!!!

I have uploaded the entire MAD libs, including the project files, zlib 1.1.4 and the winamp plugin code. You can download it from here (1.76Mb): http://homepage.ntlworld.com/jfe1205/MAD-0.15.0b.zip

to john33:
I've downloaded your MAD project, but when I opened the file "in_mad.dsw", I got a warning--told me libid3tag.dsp & libmad.dsp can not be opened.

When I compiled it, I got a lot of errors and warnings.

???

Did you try compiling the project yourself?

Ooops! Sorry, the project files worked here but they weren't correct. I've uploaded corrected ones and you can get them here: http://homepage.ntlworld.com/jfe1205/in_mad-prj.zip

The project files for in_mad have been replaced in the full download as well.

@john33:
does your plugin support the new metadata system in WA 1.9x???

X-Fixer worked out to add some (read only so far) new-metadata support in his in_flac plugin and told he would share his knowledge. maybe PM him???

This is NOT my plugin!! This is the plugin code provided with 0.14.2b simply recompiled with the 0.15.0b libs. I'm not averse to looking at the code, but I can't promise when I may get round to it.

Ciao..

> is there any quality changes in mad..
over 0.14.2b...
ex. dithering.

> what is the *best* possible command line
be for wav/cdda..
(or is the default.. *best*..)

John33, I've downloaded your latest project file.

I've compiled it OK under "Release" Mode, but I got a lot of errors and warnings under "Debug" Mode.

Did you compile it under "Debug" Mode yourself? Can you fix it and upload a newer project?

Nope, I haven't compiled under debug mode. I'll fix the project file and upload later today.

OK, I've updated the .dsp project file for compiling in debug mode, and it does compile cleanly! It's uploaded and available here: http://homepage.ntlworld.com/jfe1205/in_mad-prj.zip

I also updated it in the full MAD source download.

OK, Now! THX for your great job.
I've released my new QCDMAD.dll--a MAD input plug-in for QCD.
Anyone can download it here:
http://www.quinnware.com/plug_in1.html

BTW: to john33
Someone said you used Intel (ICL). Can you tell me what is it? A compiler or a development tool or others...
Sorry for my stupid Q

ICL is the Intel C/C++ Compiler
http://www.intel.com/software/products/compilers/cwin/

It's usually the compiler that generates fastest object code for x86 architectures.

OK, here are my MAD builds using GCC 3.2.3, including the winamp plugin, in_mad.dll:

madplay (default) (135KB)

madplay (optimized for accuracy) (135KB)

madplay (optimized for speed) (135KB)

MAD plugin 0.15.0b (default) (122KB)

MAD plugin 0.15.0b (optimized for accuracy) (122KB)

MAD plugin 0.15.0b (optimized for speed) (122KB)

Here is a screenshot for the MAD plugin.

Enjoy.

Edit: A link to the screenshot instead of an image, added size to the links.

So we've a lot of options now... stop the compilers!!!
Which is the best? Is there a true difference between compilers and between "optimized for accuracy - speed"?

The VC6 compiles that I produced are based on the default setting. If anyone wants compiles with the other settings, just ask.

You be the judge.

I personally use foobar2000. B)