On Oct 6, 2005, at 10:44 AM, Allen Briggs wrote:

> On Thu, Oct 06, 2005 at 04:24:38PM +0900, YAMAMOTO Takashi wrote:
>> if no one objects, i'll commit the following patch.
>
> So effectively, if the child has accumulated more estimated cpu time,
> the parent takes that on, but no more until it executes more.  This
> makes sense to me for most cases.
>
> The case where I think it does not make sense is if a process forks
> a few times, does some work and consumes a lot of CPU resources.
> The existing method will intentionally charge the sum of that usage
> to the parent, causing a fairly strong feedback to slow the parent
> (and newly spawned children) down.  I'm not sure that this is
> sufficient justification for keeping things the way they are (it
> seems a less common case and I don't know if this feedback is even
> really useful there).

That specifically was the intent of the change.  Please discuss with  
Ross Harvey.

-- thorpej