Re: [SCHED] wrong priority calc - SIMPLE test case

[Con Kolivas <kernel@kolivas.org>]

On Friday 13 January 2006 12:13, Con Kolivas wrote:
> On Saturday 31 December 2005 00:52, Paolo Ornati wrote:
> > WAS: [SCHED] Totally WRONG prority calculation with specific test-case
> > (since 2.6.10-bk12)
> > http://lkml.org/lkml/2005/12/27/114/index.html
> >
> > On Wed, 28 Dec 2005 10:26:58 +1100
> >
> > Con Kolivas <kernel@kolivas.org> wrote:
> > > The issue is that the scheduler interactivity estimator is a state
> > > machine and can be fooled to some degree, and a cpu intensive task that
> > > just happens to sleep a little bit gets significantly better priority
> > > than one that is fully cpu bound all the time. Reverting that change is
> > > not a solution because it can still be fooled by the same process
> > > sleeping lots for a few seconds or so at startup and then changing to
> > > the cpu mostly-sleeping slightly behaviour. This "fluctuating"
> > > behaviour is in my opinion worse which is why I removed it.
> >
> > Trying to find a "as simple as possible" test case for this problem
> > (that I consider a BUG in priority calculation) I've come up with this
> > very simple program:
>
> Hi Paolo.
>
> Can you try the following patch on 2.6.15 please? I'm interested in how
> adversely this affects interactive performance as well as whether it helps
> your test case.

I should make it clear. This patch _will_ adversely affect interactivity 
because your test case desires that I/O bound tasks get higher priority, and 
this patch will do that. This means that I/O bound tasks will be more 
noticeable now. The question is how much do we trade off one for the other. 
We almost certainly are biased a little too much on the interactive side on 
the mainline kernel at the moment.

Cheers,
Con
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/