sched_autogroup interactivity patch for the desktop
jwboyer at gmail.com
Wed Nov 17 11:31:27 UTC 2010
On Wed, Nov 17, 2010 at 5:03 AM, Ilyes Gouta <ilyes.gouta at gmail.com> wrote:
> I've been reading http://lkml.org/lkml/2010/11/16/402 and actually Lennart
> has put some solid thoughts and arguments there and IMHO it was *somehow*
> pretty convincing that this should be done at user-space, in conjunction
> with tools such as systemd, albeit introducing more complexity in form of
> configuration infrastructure and whatnot.
> Having this being done automatically by the kernel looks really easy and yet
> promising. The downside of it is that it just feels like it's something
> rigged towards a special class of applications (TTY based, build jobs) and
> would act behind the scene without user-space knowing (explicitly) about
> this classification, cgroups being spawned and subtasks associated to them.
> Still, It would be nice to have this differentiation in place, being
> implemented in either forms, since it brings a certain win for the desktop
> user and developer. I think having a term open for churning code is part (or
> a possible user-case) of the desktop experience :)
Being entirely serious, you should have sent this email on the thread
you indicated. Sending it to this list isn't going to really benefit
the discussion that needs to happen.
More information about the devel