On Mar 28, 2007, at 10:32 AM, Jason Thorpe wrote:

>
> On Mar 28, 2007, at 6:40 AM, <dsieger@techfak.uni-bielefeld.de> wrote:
>
>> On Wed, Mar 28, 2007 at 09:48:04PM +0900, YAMAMOTO Takashi wrote:
>>>
>>> a possible way is what recent solaris does.
>>> http://blogs.sun.com/eschrock/date/20041013
>>
>> Interesting read. Thanks for the link. However, I think this  
>> should be
>> dealt with in a separate project (maybe add a comment to src/doc/TODO
>> or the projects list?) and should not hinder us from going ahead with
>> the proposed split.
>
> I think that's fair... however, I would consider it a pretty high- 
> priority item to be addressed, because the current situation  
> hinders the scalability of the system.

I have a set of changes somewhere for powerpc to get/save timebase on
interrupt, exception, idle, and cpu-switch for precise measurements
of idle, user, system, and interrupt (and per-lwp) time.

This was part of getting away from a fixed hardclock interrupt.

I probably can dig them out or resurrect them easily.