Re: CFS Scheduler and real-time tasks

On 5/19/07, Török Edvin <[EMAIL PROTECTED]> wrote: I tried -v13. However the scheduling "error" is now 10% (vs 2% with -v12). I also noticed strange behaviour with CPU hotplug. I offlined cpu1 (echo 0 >/sys/devices/system/cpu/cpu1/online), and the typing speed on my terminal decreased noticably.

Re: CFS Scheduler and real-time tasks

On 5/17/07, Ingo Molnar <[EMAIL PROTECTED]> wrote: hey, cool! Also try -v13 - it should be even more tighter. I tried -v13. However the scheduling "error" is now 10% (vs 2% with -v12). I also noticed strange behaviour with CPU hotplug. I offlined cpu1 (echo 0 >/sys/devices/system/cpu/cpu1/onli

Re: CFS Scheduler and real-time tasks

* Török Edvin <[EMAIL PROTECTED]> wrote: > On 5/17/07, Ingo Molnar <[EMAIL PROTECTED]> wrote: > > > >note: you ran it on a dual-core system and the SMP load-balancer needs > >time to gain precision in the non-RT test - so i'd suggest to test it > >much longer than 10 seconds: 300 seconds at least

Re: CFS Scheduler and real-time tasks

On 5/17/07, Ingo Molnar <[EMAIL PROTECTED]> wrote: note: you ran it on a dual-core system and the SMP load-balancer needs time to gain precision in the non-RT test - so i'd suggest to test it much longer than 10 seconds: 300 seconds at least. Thanks, running for 300 seconds shows only a 2% err

Re: CFS Scheduler and real-time tasks

* Török Edvin <[EMAIL PROTECTED]> wrote: > Is there something I am doing wrong, or is the precision decrease > expected, when running real-time tasks too? yeah: real-time tasks are ueber-tasks that just get all CPU time immediately. I guess we could make SCHED_RR be covered by CFS accounting