[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CABcS5h0tyDjmFSg+sUeJ14m8KsTxNhwo=qLHFLDX8J-tCpdDOw@mail.gmail.com>
Date: Mon, 1 Aug 2011 11:45:12 +0100
From: Rolando Martins <rolando.martins@...il.com>
To: Mike Galbraith <efault@....de>
Cc: Peter Zijlstra <peterz@...radead.org>,
Thomas Gleixner <tglx@...utronix.de>,
LKML <linux-kernel@...r.kernel.org>,
linux-rt-users <linux-rt-users@...r.kernel.org>,
"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
Subject: Re: [ANNOUNCE] 3.0-rt6
On Mon, Aug 1, 2011 at 11:34 AM, Mike Galbraith <efault@....de> wrote:
> On Mon, 2011-08-01 at 09:42 +0100, Rolando Martins wrote:
>> Hi Peter,
>> is there any progress on the CONFIG_RT_GROUP_SCHED front?
>
> Out of curiosity, why would you want that in a -rt kernel?
I am building a rt middleware that uses cgroups to reserve cpu
bandwidth for a set of tasks.
But I also interested in minimization latencies...
>
> (I haven't done any PI testing with that, but I can imagine the throttle
> causing heartburn)
>
> -Mike
>
>> I don't know if this is related, but in 33-rt if I used a low (=100)
>> cpu_runtime_us and cpu_period_us then I would get some freezes
>> (something that does not happen in 2.6.39).
>
> (providing a realtime budget of 100 whole microseconds to a kernel where
> everything and it's brother is realtime is unlikely to go well)
I empirically tested values until I reached this one. Anything above
this would result in a huge amount of jitter
(to all tasks under a control group), at least on my boxes (corei7
920, and AMD phenom x4 950).
Rolando
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists