to thread the X server (?)
vignatti at c3sl.ufpr.br
Tue Jun 24 20:39:25 PDT 2008
Xavier Bestel escreveu:
> On Wed, 2008-06-18 at 23:14 +0200, Daniel Stone wrote:
>> On Wed, Jun 18, 2008 at 10:56:40PM +0200, Juliusz Chroboczek wrote:
>>> I appear to have missed part of the discussion. Why are you trying to
>>> multithread the X server? It will cause a lot of subtle bugs, and
>>> I don't see what it will buy you.
>> One interesting path I see with the input-in-a-separate-thread work is
>> eventually getting a complete path from hardware input event to client
>> delivery that cannot be blocked by rendering, meaning we always have
>> very low latency on input events.
> Xorg has a problem latency-wise: clients can block each other. When I
> play a video (in Totem), and simultaneously use a web browser, the video
> skips a lot - on a dual-proc system.
> I don't think just separating the input processing from the client
> requests processing will solve that.
The interesting thing (or not so interesting) is if we set the old
scheduler  it has a better experience than the "smart" scheduler.
Well at least here in my machine using XVideo extension to play videos.
Seems that timeslice introduced by smart scheduler isn't helping so much
when multiple clients are running.
 try yourself: ./Xorg -dumbSched & mplayer -vo xv something.avi &
C3SL - Centro de Computação Científica e Software Livre
More information about the xorg