Re: [RFC PATCH] Dynamic sched domains (v0.5)

[Dinakar Guniguntala <dino@in.ibm.com>]

On Tue, May 03, 2005 at 03:03:23PM -0700, Matthew Dobson wrote:
> An interesting feature.  I tried a while ago to get cpusets and
> sched_domains to play nice (nicer?) and didn't have much luck.  It seems
> you're taking a better approach, with smaller patches.  Good luck!

Thanks ! I would very much like to know your findings as far as
memory/node domains are concerned or are you going to be working on it?
I dont have any thoughts on it right now

> > -#ifdef CONFIG_HOTPLUG
> > +#if defined(CONFIG_HOTPLUG) || defined(CONFIG_CPUSETS)
> >  #define __devinit
> >  #define __devinitdata
> >  #define __devexit
> 
> This looks just plain wrong.  Why do you need this?  It doesn't seem that
> arch_init_sched_domains() and/or update_sched_domains() are called from
> anywhere that is cpuset related, so why the #ifdef CONFIG_CPUSETS?

cpu_attach_domain is defined as a __devinit, maybe I need to remove that
instead of the #ifdef

> >  #ifdef CONFIG_SMP
> > -#define SCHED_DOMAIN_DEBUG
> > +#undef SCHED_DOMAIN_DEBUG
> >  #ifdef SCHED_DOMAIN_DEBUG
> >  static void sched_domain_debug(struct sched_domain *sd, int cpu)
> >  {
> 
> Is this just to quiet boot for your testing?  Is there are better reason
> you're turning this off?  It seems unrelated to the rest of your patch.
> 

This gets called from cpu_attach_domain, and so everytime partitioning is done
and not only during boot with my changes

Thanks for your review !

	-Dinakar