[Mono-list] csvorbis in CVS

[Mark Crichton]

--DocE+STaALJfprDB
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Hello,

After much bitbanging, poking around in the mono JIT, bugging people,
and doing some code edits, I would like to announce that the C#
implementation of Ogg Vorbis is now in mono CVS.

The module is csvorbis.  I've been using VS.NET and the MS runtime to
compile the code, so I don't know if it will compile with mcs (it
should however).  It does run, however, with the mono JIT and class
libs.  The performance of the code still needs to be tested, but on the
mono runtime on my large hammer of a computer (1900+ MP class box), I
can decode a 6 minute ogg in about 50 seconds.  Other data points would
be nice... (which raises a question, does the x86 JIT emit floating
point instructions?)

The code is far from complete, however, there are a number of things to
be done:

-  I just have project files for VS.NET.  If anyone is familiar with
nant, I'd appreciate it if people could do some build files for it.

-  Casting.  I've done a lot of it.  Some of the functions should be
cleaned up with that regard, to make them look more like their C
counterparts.

-  Threading issues.  If I came across a synchronization() block in the
Jorbis code...I just kinda ignored it. :)  That needs to be fixed.

-  internal methods/variables.  Java and C# are different in regard to
default accessibility.  So, a lot of the code has internal methods
tagged on.  I think many of them should be cleaned up in some way,
I just went through whole classes and mass-marked functions internal.
Some of the classes might need to be cleaned up in that regard.  The
Ogg parser (csogg) is really bad in this accord.

-  Math errors.  There's some math error somewhere in the code.  The
resulting raw PCM seems tohave the lowest order bit missing/set in some
cases.  This doesn't result in a easily detectable audio error (it only
appears for one sample at most), however it does spew output that is
different from the reference code.  This is a bug, but I have no idea
where it is coming from.

-  Documentation.  This is a necessity.  The sample code in
OggDecode.cs shows the general flow of how to use the decoder.

-  TODO items.  They're marked in the code.  Mostly converting Ogg
comments (byte arrays) into Strings.  Shouldn't be too hard.

Wishlist items:

-  Checking to see if the decoder has changed from when Jorbis was done
to what 1.0rc3 does now.  I doubt there is a difference, but it should
be checked.

-  The classes could be cleaner.  See internal bit above.

-  Encoder work.  Harder to do, since someone will need to convert the
codebooks over.  Look at the libvorbis code for that.

-  Gtk# demo program.  This would rock for Miguel's demoing of mono.

If anyone is interested in the above, or has any other ideas on how to
(ab)use the code, let me know.=20

Thanks,
Mark Crichton=20

--DocE+STaALJfprDB
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (SunOS)
Comment: For info see http://www.gnupg.org

iEYEARECAAYFAjzMH00ACgkQOfj2Ja/u/oAf+wCeN3V9gLoqeghtZAXywB7gSOWs
T7YAni9OK5VEeTXgisqcF1TLjOoNRNC4
=P0cL
-----END PGP SIGNATURE-----

--DocE+STaALJfprDB--