[MPlayer-dev-eng] [PATCH] liba52 float support

[D Richard Felker III]

On Sun, Feb 20, 2005 at 02:12:10PM -0600, Zoltan Hidvegi wrote:
> Reimar Döffinger wrote:
> > Hi,
> > On Sun, Feb 20, 2005 at 02:13:35PM -0500, D Richard Felker III wrote:
> > > On Sat, Feb 19, 2005 at 07:27:18PM -0600, Zoltan Hidvegi wrote:
> > > > And if float is not broken, should a52float be the default?  There is
> > > > almost no speed penalty, as float is the native format of liba52
> > > > anyways.
> > > 
> > > It definitely IS broken. Before you make it default you need to fix
> > > libaf float support so it works...
> > 
> > Can you give some details? This really is _not_ helpful for fixing. If
> > you can't provide the info I'm for just committing and waiting for some
> > ennerved users to help us fix it.
> 
> Do you mean commiting with float as default?  But yes, I've like to
> know more details about what's broken.  Maybe it's broken on 3Dnow, I
> think Rich still uses a K6.

If any float processing gets done in libaf, it loads several
conversion filters to convert back and forth between many different
nonsense formats, generating 8bit samples in the end. And somewhere
during all the processing, the audio becomes nothing but white noise.
This has always happened ever since libaf was first written.

Yes, I use a K6, so it may have something to with with 3dnow, and/or
missing emms/femms between mmx/3dnow and float usage. I'll try it
again soon and post logs, but you had better believe this bug really
exists and anyone committing code that makes the broken libaf float
crap get used by default will be drinking 1000000000l of pepsi unless
they find and fix the bug first!

Rich