Re: 2.6.8-rc2-mm2

From: Nick Piggin
Date: Wed Aug 04 2004 - 06:13:31 EST


Con Kolivas wrote:
Nick Piggin writes:

And child-runs-first in wake_up_new_task. Please don't.


It does child runs first by design in staircase. You don't need any more.


Oh ok. Well please don't remove the check anyway.

Also, basic interactivity in X is bad with the interactive sysctl set to 0


Well duh... disable interactivity and interactivity is bad. What's the problem? It's not meant to be used on a desktop in that way.


Well why would you want to disable it then?

(is X supposed to be at nice 0?), however fairness is bad when interactive is 1.
I'm not sure if this is an acceptable tradeoff - are you planning to fix it?


Why? A single user desktop is hardly needing extremely accurate cpu distribution... we see that already in 2.6.


No, I'm talking about one task getting *double* the amount of CPU as
another, when it should actually be getting a little bit less.

It has interactivity problems with "thud". Also the mouse can freeze for .5 to 1
second when moving between windows while there is disk IO going on in the background
(this is with interactive = 1). The test-starve problem is back.


Hmm? a minor mouse freeze with a _test_ starvation program is not starvation; nor is it an interactivity problem. Yours is the first complaint about interactivity during i/o.

The mouse was often freezing for me when moving over windows with
IO running. That was an interactivity problem for me.

Thud itself isn't a problem I guess. Just thought you might like to know.

The test-starve problem actually is a problem though.


Increasing priority (negative nice) doesn't have much impact. -20 CPU hog only gets
about double the CPU of a 0 priority CPU hog and only about 120% the CPU time of a
nice -10 hog.


-20 is 40 rr intervals. 0 is 20 rr intervals. +19 is 1 rr interval. Seems to me the cpu distribution is working our absolutely perfectly as designed.


Well maybe it should be re designed. If an admin wants a task to get a
decent amount of CPU, it is not reasonable to nice +19 everything else.

Why is the only critic of this the person with a competing design? Does anyone else object to these things? I certainly dont feel objective enough to criticise yours.


No need to get a bee in your bonet. I was pointing out some things that
you need to fix if you are serious about this. I'd rather not replace
the scheduler with a known broken one.

Oh, and I would very much like you to criticise mine :)
-
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/