Re: Schedulers benchmark - Was: [ANNOUNCE][RFC] PlugSched-5.2.4 for 2.6.12 and 2.6.13-rc6

From: Con Kolivas
Date: Fri Aug 19 2005 - 19:32:42 EST


On Sat, 20 Aug 2005 06:13, Lee Revell wrote:
> On Fri, 2005-08-19 at 14:36 +1000, Con Kolivas wrote:
> > On Fri, 19 Aug 2005 02:41 pm, Peter Williams wrote:
> > > Maybe we could use interbench to find a nice value for X that doesn't
> > > destroy Audio and Video? The results that I just posted for
> > > spa_no_frills with X reniced to -10 suggest that the other schedulers
> > > could cope with something closer to zero.
> >
> > I don't see the point. X works fine as is without renicing not
> > withstanding these extreme loads in interbench. Furthermore, reworking of
> > xorg code to not spin the cpu unnecessarily when the gpu is busy is
> > underway and tuning the cpu scheduler unfairly for an X server that will
> > no longer behave so badly is inappropriate.
>
> See, that's where we disagree, I certainly don't believe X "works fine".
> Compared to MacOS and (especially) Windows the Linux desktop is WAY
> sluggish.
>
> For example when I cycle through windows with alt-tab in X, it can take
> 5-10 seconds for each to render. I can see the application's widgets
> being drawn one at a time, then finally the border. Repeated
> alt-tabbing between the same two windows seems to cause a CPU intensive
> redraw of the entire window. It's as if X just discards the rendered
> contents of a window as soon as it's obscured.
>
> On Windows this works as expected - cycling through windows whose
> contents have already been rendered is *instantaneous*.
>
> I agree that tweaking the scheduler is probably pointless, as long as X
> is burning gazillions of CPU cycles redrawing things that don't need to
> be redrawn.
>
> Then again even the OSX scheduler has hooks for the GUI. Presumably
> they concluded that the desktop responsiveness problem could not be
> adequately solved within the framework of a general purpose UNIX
> scheduler.

It's an X problem and it's being fixed. Get over it, we're not tuning the
scheduler for a broken app.

Con
-
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/