comments on KDE performance tips

Alexander Kellett kde-optimize@mail.kde.org
Mon, 13 Jan 2003 15:02:03 +0100


On Mon, Jan 13, 2003 at 03:20:22PM +0100, Mathieu Kooiman wrote:
> Ah, well, I just skimmed over the names on KernelTrap.. :)  

:)

> I know that 'responsiveness' is crucial to desktop performance (or the user 
> experience thereof) but isn't KDE spending most of its time in "user land" 
> (in kernel terms) ?

yes, but its not the time its taking thats important, its which process
is getting the "time". and that is upto the kernel to decide. (i'm including
system time in this)

> The preempt patch doesn't exactly speed that up. AFAIK, the preempt patch 
> introduces more and more places in supposably critical code sections in the 
> kernel where the kernelprocess can be preempted. 

basically from what i understand, if you start a program yet continue to
work in another (for example you load mozilla using minicli and then wish
to use an already opened konsole) a preempt kernel will allow the keystrokes
to quickly be processed, whereas a kernel without preempt would in all
likelyhood like to load all of mozilla up into memory (i.e disk i/o which 
is a kernel job of course) before beginning to process the keystrokes.
low-latency does basically the same thing from what i understand, but in
a more brute-force way. usually it produces better results though and preempt
in 2.4 kernel isn't entirely perfect from what i understand.

> Then again, I'm by no means claim to be an expert on these things :)

nor me :)

Alex

-- 
"[...] Konqueror open source project. Weighing in at less than
            one tenth the size of another open source renderer"
Apple,  Jan 2003 (http://www.apple.com/safari/)