Todd's affinity code

[Arne Chr. Jorgensen]

Todd Denniston wrote:
> Message: 10
> Date: Mon, 20 Apr 2009 19:01:59 -0400
> From: Todd Denniston <Todd.Denniston at ssa.crane.navy.mil>
> Subject: Re: Fedora-10 nightmare update2
> To: For testers of Fedora Core development releases
>    <fedora-test-list at redhat.com>
> Message-ID: <49ECFEE7.5060802 at ssa.crane.navy.mil>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> <SNIP>
>
> Question:
> if you do
> taskset -p 0x00000002 \
>  `ps aux |grep /usr/bin/Xorg| \
>   grep -v grep |awk '{print $2}'`
>
> {painful way to get the X pid to feed to taskset, which suggests to the kernel
> keep the X process on the second processor.}
>
> and then run your tests, does the problem seem less troublesome at the lower
> speeds?
>
>-- 
> Todd Denniston
> Crane Division, Naval Surface Warfare Center (NSWC Crane)
> Harnessing the Power of Technology for the Warfighter
>

"..painful way to get the...", - not when you wrote it. But to compose anything
like that oneself, - that is painful ;)

But it did not improve.   

pid 3467's current affinity mask: 3
pid 3467's new affinity mask: 2

( did try it the other way around as one core has a higher temperature then the other )
..think the load was 5% on one, and 30% on the other.. approx.

At 800MHz, there are pauses of several seconds. Changed the 'time of day' to show seconds, and it passed a few numbers as it caught up again. 

What I don't understand is there are many other machines like this, - is there something
broken with my hardware ?  But in 'top', it is somewhat clear that Xorg is eating time..

Thanks Todd, your code worked as a charm. The others may have missed one of those ''

//ARNE