Re: init and USB oddities-ULE-ATA

On Monday 17 November 2003 06:08, Steve Kargl wrote:
> On Mon, Nov 17, 2003 at 05:41:04AM +0100, Harald Schmalzbauer wrote:
> Content-Description: signed data
>
> > On Monday 17 November 2003 05:25, Steve Kargl wrote:
> > > On Mon, Nov 17, 2003 at 04:39:08AM +0100, Harald Schmalzbauer wrote:
> > > Content-Description: signed data
> > >
> > > > Next I'd like to report is what I already mentioned in "ULE and very
> > > > bad responsiveness"   I followed Jeff Roberson hint and ran
> > > > setiathome with nice 20. But this didn't really change anything.
> > >
> > > ULE has been rock solid for me since Jeff's last
> > > major update.  Of course, I run neither setiathome
> > > nor KDE.
> >
> > Give setiathome a try! You'll be astonished.
>
> No thanks.  It's a waste of CPU cycle.

Well, certainly you're rigth. But I have spare cycles and with 4BSD scheduler 
they were handled very well.

>
> > And I'm sure the difference I _feel_ isn't dependend on kde. If you
> > don't like kde replace it with our favourite wm/desktop.
>
> I prefer fvwm2.  ULE works fairly well.
>
> > But you won't be able to play two mid to high-quality
> > mpegs at the same time on a 1GHz machine where 4BSD scheduler does very
> > well!
>
> I can assure you that the numerical simulations I run, along with
> the "make worlds", and compilations of gcc's tree-ssa branch
> stress the system.  I re-install over 100 ports today and the
> load average was rarely below 5.  I was use linux-opera and
> knews and sylpheed and several other program and noticed
> nor degradation in responsiveness.  Does seti cause a problem
> if you are not running X (or KDE).

Yes. The difference is the same. Like I originally mentioned (on one single 
cons25) with seti in the background (doesn't matter if nice is 15 or 20) it's 
almost impossible to wait until a "make clean" of a port with little 
dependencies (like nvidia-driver) finishes.

>
> > I haven't claimed ULE to be unstable though. I just wanted to highlight
> > some issues which will be a big problem if 5.2-releas will have ULE as
> > default!
>
> If ULE is destined to be the default scheduler in 5-stable, then
> we need to have more people test it.

I can copy that. That's the reason why I started to try ULE. The notes in the 
kernel didn't convince me really;) It was a thread in a newsgroup and after 
posting my problems Kris told me to post in -current.

Best regards,

-Harry