The Linux desktop is about to get a LOT faster

[Christopher Chan]

On Friday, November 19, 2010 11:04 PM, Samuel Thurston wrote:
> On Thu, Nov 18, 2010 at 11:35 PM, Christopher Chan
> <christopher.chan at bradbury.edu.hk>  wrote:
>> HOWEVER
>>
>> http://www.webupd8.org/2010/11/alternative-to-200-lines-kernel-patch.html
>>
>> Redhat chap quickly showed a 'work around'. methinks i should be running
>> Fedora/Centos.
>
> A) there are separate (untested) instructions for applying the same
> workaround to Ubuntu in that article as well as two patched
> Ubuntu-specific kernels.  Don't go fleeing to redhat just yet :)

Fleeing? Ahem, I am visiting from Redhat/Centos and from OpenSolaris.


>
> B) something seems odd about the performance testing numbers posted to
> the lkml.  I can't put my finger on it and don't have the time to run
> proper benchmarks of my own at the moment.  This weekend perhaps I'll
> add the Lennart method on my home machine and see how it feels. What
> concerns me is that because the cgroups are tty-bound you won't see
> performance increases on system services and things like networking
> (for which the benchmark numbers do not account.) that's just a hunch
> though.
>

Oh, I was only thinking about the desktop. I did not notice that system 
services were implicated too in the 233 line patch.