Re: Linux 2.6.0-test6

[Nick Piggin]

Ed Sweetman wrote:

> Nick Piggin wrote:
>
> > Rob Landley wrote:
> >
> > > On Sunday 28 September 2003 02:03, Con Kolivas wrote:
> > >
> > > > On Sun, 28 Sep 2003 11:27, Linus Torvalds wrote:
> > > >
> > > > > from Andrew Morton. Most notably perhaps Con's scheduler changes that
> > > > > have been discussed extensively and made it into the -mm tree for
> > > > > testing.
> > > > For those who are trying this for the first time, please note that the
> > > > scheduler has been tuned to tell the difference between tasks of 
> > > > the _same_
> > > > nice level. This means do NOT renice X or it will make audio skip 
> > > > unless
> > > > you also renice your audio application by the same amount. Lots of
> > > > distributions have done this for the old 2.4 scheduler which could not
> > > > treat equal "nice" levels as differently as the new scheduler does 
> > > > and 2.6
> > > > shouldn't need special treatment.
> > > >
> > > > So for testing note the following points:
> > > >
> > > > Make sure X is NOT reniced to -10 as many distributions are doing.
> > > > Some shells spawn processes at nice +5 by default and this will 
> > > > make audio
> > > > apps suffer.
> > > > Make sure your hard disk, graphics card and audio card are 
> > > > performing at
> > > > equal standard to your 2.4 kernel (ie dma is working, graphics is 
> > > > fully
> > > > accelerated etc).
> > >
> > > I.E. with your new scheduler, priority levels actually have enough 
> > > of an effect now that things that aren't reniced can be noticeably 
> > > starved by things that are.
> >
> > AFAIK, Con's scheduler doesn't change the nice implementation at all.
> > Possibly some of his changes amplify its problems, or, more likely they
> > remove most other scheduler problems leaving this one noticable.
> >
> > If X is running at -20, and xmms at +19, xmms is supposed to still get
> > 5% of the CPU. Should be enough to run fine. Unfortunately this is
> > achieved by giving X very large timeslices, so xmms's scheduling latency
> > becomes large. The interactivity bonuses don't help, either.
>
> there are 40 positions between -20 and 19, that doesn't equal 5% 
> steps. They


No, but the maximum timeslice (sole metric changed by nice) is 200, the
min is 10 (5%). And between them, timeslices are calculated linearly.

> don't even refer to % of cpu.  If i nice a process to -20 it doesn't 
> get a given percentage of cpu just because it's -20. I may have other


No, but it should get 2000% of what a nice +19 process will get in the
same system (regardless of what else is running).

> processes at -20 as well.  If you nice something to -20 and it is 
> actually using that cpu then things that are +19 shouldn't run and


They do.

> wont run.  If I nice -20 vmstat 1, it's not going to starve xmms (or 
> any better audio player).  -20 means starve all and it should do that 
> when it actually makes use of the resources. 


I don't know exactly what nice is supposed to do other than "raise
priority", but its fairly well accepted that it should increase the
process' % cpu time (vs others) without completely starving everyone.

It is probably a justified criticism that 5% is too much for a +19 task
to get vs a -20 task, but it has to get something.

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