Audio output 3 status & call for contributors

Christophe Massiot massiot at
Mon Aug 12 00:51:34 CEST 2002

At 15:02 +0200 10/08/2002, Jean-Paul Saman wrote :

>>- libmad
>Starting on libmad ... do the downscaling functions need to move to 
>"fixed32 resampler" or another audio filter ?

Yes you will need to write two converters. The mad codec should 
request buffers of type AOUT_FMT_FIXED32 (vlc_fixed_t is defined to 
be strictly identical to mad_fixed_t), and copy raw samples from 

Then two things may happen : either the host has a FPU, and the aout 
core will spawn a float32 mixer, and in that case will look for a 
fixed32tofloat32 converter module.

Or the host has no FPU, and the aout core will spawn a fixed32 mixer. 
The trivial mixer can also mix fixed32 samples, so that won't be a 
problem. However, the output plug-in will require a fixed32tos16 
module on most architectures (ie. all except Mac OS X).

You can also provide different converters with different 
precision/efficiency, but currently there is no way to choose one at 

Christophe Massiot.

This is the vlc-devel mailing-list, see
To unsubscribe, please read
If you are in trouble, please contact <postmaster at>

More information about the vlc-devel mailing list