[Scummvm-devel] SCI sound drivers

[Walter van Niftrik]

Max Horn wrote:
> Right now, the polled driver works fine, but if we could get  
> everything to work with the realtime driver (without regressions, mind  
> you), that would of course be nice: Then there would be no need for  
> switching, and a single code path to test.
>   

In this case I would consider "less accurate audio timing" a regression 
as well. In any case, both the FreeSCI stable branch and the GSoC 
ScummVM code perform a lot worse in this regard than the polled player. 
The question is whether this regression can be avoided.

Regards,

Walter