CVS Commit: vlc

[Laurent Aimar]

On Wed, May 01, 2002, Jean-Paul Saman wrote:
> In the mad_libmad.c code there are two downscale algorithmes. One of 
> them is adaptive on dither errors and masks rough edge from the sound. 
> You could try to use this function instead of the default *faster* 
> downscaling.
> 
> In fact it is an algorithme borrowed from mpg321
> //#define MPG321_ROUTINES     1
> uncomment it and try again. Check the value of AOUT_BUFFER_DURATION it 
> may also influence your result.
> 

  I've tried but it make no amelioration at all.

 I'v searched and (thanks to added intf_Msg ...) i'v found that with
the mpeg audio streams that are awfull, mad function don't return
ADEC_FRAME_SIZE(2*1152) but half of it.

 So when, libmad_output received these data, it fill a buffer of
ADEC_FRAME_SIZE long with half of real data and then switch to the next
data buffer by increase p_mad_adec->p_aout_fifo->i_end_frame .But then
half of this buffer is not used and let blank.

 I think the way to solve it is to keep the number of samples that are
out, and so switch to next data buffer only when is really needed . Or
find a way to known what sort of stream make mad to not output the good
size, and create fifo with good size.

-- 
fenrir


-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://www.videolan.org/lists.html
If you are in trouble, please contact <postmaster at videolan.org>