Re: Desktop Kernel Stuff

[Alan Cox <alan redhat com>]

> Murray: In fact, I think 2.5 does have scheduler changes that
> distributes spare processor time from X clients up to the X server if
> the clients don't actually need it. That starts to recognise X as being
> an "interactive" process due to all its interactive clients instead of a
> background daemon-type process. And so, it gets more processor time if
> its available.

X generally shows up as interactive because it sleeps a lot. We don't 
actually go as far as some non Unix systems where you can pass the 
rest of your timeslice to someone (so X for example could say 
"Im stuck waiting for X to do things on my behalf, give X my cpu quota"

> But that explanation could be really wrong, I am not a kernel hacker.

Reduced to its basics schedulers try and do something like

	priority = recent_idle_time/recent_total_time;

so the less CPU you use the quicker you get the CPU but if you hog it
you will rapidly lose it to other tasks as your priority drops.