[vlc-devel] [PATCH] aout: pulse: use jitter delay

Thomas Guillem thomas at gllm.fr
Wed Feb 6 09:10:21 CET 2019


On Tue, Feb 5, 2019, at 17:21, Rémi Denis-Courmont wrote:
> Le tiistaina 5. helmikuuta 2019, 15.04.07 EET Thomas Guillem a écrit :
> > Yes, it's better to use the period as an IN/OUT parameter and let the core
> > calculate the jitter from it.
> 
> I don't think that messing with audio buffering before the input clock rework 
> is good or even workable idea. The current values constitute a tradeoff that 
> mostly works with the current input buffering - it's unlikely that changing it 
> will do any good at this point.

Yes,  we should keep the same period/buffering for 4.0. But I would to do some tests, specially on macOS. If this current API and new clock solve airplay playback, it is a very thing to have (mac os aout module telling the core to buffer for 2seconds then).

> 
> But in any case, there is no way that the aout can return a jitter value. 
> Typically, audio works with a period time (or period sample count) and a buffer 
> time or buffer size. But that's just how hardware most commonly works - I don't 
> think you can assume and expose that in the abstraction.
> 
> And to compute the correct buffering values, the aout needs to know the 
> packetization time, the maximum buffer duration (if any) and what to optimize 
> for (low latency or high reliability and low power).
> 
> -- 
> Реми Дёни-Курмон
> http://www.remlab.net/
> 
> 
> 
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel


More information about the vlc-devel mailing list