Re: linux-next: concerns over the perfmon3 tree

["stephane eranian" <eranian@googlemail.com>]

Ingo,

On Wed, Nov 26, 2008 at 11:47 AM, Ingo Molnar <mingo@elte.hu> wrote:
>
> * Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
>> Hi Stephane,
>>
>> Given the concerns expressed by Ingo, I have dropped the perfmon3
>> tree from linux-next until they have been addressed.  Hopefully that
>> will not be loo long.  Let us all know how we can help with this.
>
> Thanks, we'll try to finish the review ASAP (today+tomorrow) so that
> it can be added back in 1-2 days. And it's sensible to keep it in the
> separate perfmon3 tree as well (as long as it's not rebased) - that's
> a convenient initial structure for new features.
>

I think I am missing something about rebasing. Are you saying you want
the patches to stay with the same tree until it is reviewed thoroughly? So
far I have used Linus' v2.6.28-rcXX tree as it evolves at a pace I can follow.
Is that okay?

> ( and the powerpc bits would be nice as well, they seem to be missing
>  from the current lineup - the more architectures keep hitting on
>  that code, the better the end result. )
>

The fully featured perfmon3 code (available as a branch of my kernel.org tree)
has all the architectures we currently support. There is some overlap with the
patchset posted on LKML. I focused on x86 because this is the key architecture
with a lot of demands. It should not be very difficult to derive the
powerpc patchset
from the fully featured git branch, same goes for MIPS, and SPARC. IA-64 will be
a bit more challenging because they already have a much older version
of perfmon2.
Switching IA-64 to perfmon3 now, would cause lots of perfmon2 features
to disappear
(e.g., sampling, system-wide).