On Mon, 11 Jun 2007, Srivatsa Vaddagiri wrote:
> 
> struct sched_entity
>         stores essential attributes/execution-history used by CFS core
>         to drive fairness between 'schedulable entities' (tasks, users etc)

Wouldn't this be sensible to integrate into CFS _regardless_ of anything 
else?

Especially with the alleged "pluggable" scheduler model, one of the things 
lacking in any pluggability is the fact that the core thread data 
structures are anythign but pluggable. This patch makes it no more so, but 
at least it abstracts out the fields that are used by the scheduler, and 
would seem to be a nice cleanup. No?

		Linus
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers

* Linus Torvalds <torvalds@linux-foundation.org> wrote:

> 
> 
> On Mon, 11 Jun 2007, Srivatsa Vaddagiri wrote:
> > 
> > struct sched_entity
> >         stores essential attributes/execution-history used by CFS core
> >         to drive fairness between 'schedulable entities' (tasks, users etc)
> 
> Wouldn't this be sensible to integrate into CFS _regardless_ of 
> anything else?

yeah, that's what i asked Srivatsa to do, and his currently patchset 
largely achieves that. I'm looking now into applying the core bits and 
checking that it's in essence a NOP when this isnt used. (that makes the 
later merging of group fairness and container scheduling a lot less 
contentious)

	Ingo
_______________________________________________
Containers mailing list
Containers@lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers