Re: Re: [RFC/RFT PATCH] sched: automated per tty task groups

[Marco]

I've made some tests to give a try to tty grouping patches.

Kernel 2.6.36+CONFIG_SCHED_AUTOGROUP+Improve_tick_preemption.patch

These are the result of :
taskset -c 0 ./wakeup-latency& sleep 30;killall wakeup-latency (owner marco)
vs
taskset -c 0 make -j 10 bzImage


Vanilla 2.6.36
maximum latency: 52879.7 µs
average latency: 4217.5 µs
missed timer events: 0

Kernel 2.6.36+autogroup enabled
maximum latency: 2840.8 µs
average latency: 15.5 µs
missed timer events: 0

Kernel 2.6.36+libpam-systemd enabled
maximum latency: 26361.6 µs
average latency: 4548.6 µs
missed timer events: 0

Kernel 2.6.36+START-NICE sched feature(v3)
maximum latency: 76182.8 µs
average latency: 4930.1 µs
missed timer events: 0

Quite impressive.....
Now, i can play supertuxkart while make'ing -j 10 and stellarium shows
the highest fps i've seen so far.
Very good job, guys.

Marco


On Thu, Oct 21, 2010 at 11:17 AM, Marco <mpiazza@xxxxxxxxx> wrote:
> Hello, i'm a little confused about tty grouping.
> The benefit in terms of latency is only due to the grouping of task
> under the appropriate cgroup hierarchy?
>
> Can this issue be solved by userspace daemon (think of libpam-systemd,
> libpam-cgroup)? Perhaps tty grouping in kernel space is more
> efficient.
>
> I'm missing something?
>
> Marco
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/