Re: [xine-devel] Optimum fifo sizes?

[Michael R. <ama...@fr...>]

Hi there,

> So, I increased the audio buffer count from 50 to 500 and both
> samples played.

What I'm starting here has nothing to do with this ts demuxer problem, 
but I always wanted to discuss this and I think this thread could be 
apropriate:

I always thought that the audio fifo might be too short. During dvd 
playback, which is what I want xine to be optimized for, the video and 
audio fifo are filled with blocks of 2k size, right? (not sure about 
that)
That means on a typical dvd with a video data rate of about 4 Mbit/s 
the video fifo can hold about 2 seconds of video. This is a sensible 
amount I guess and should be enough caching to compensate drive delays 
during layer changes.
But with a typical audio data rate of 1.8 MBit/s (four ac3 tracks), the 
audio fifo holds only 0.4 seconds. I think this is too small. As far as 
I understand the demuxers, this also renders the additional 1.6 seconds 
of video fifo useless, because demuxer blocks when the first fifo is 
full.
My suggestion would be to enlarge the audio fifo to hold 2 seconds too.
That would imply a size of 230 xine buffers.

Michael Roitzsch