konqy preloading kills performance

Alexander Neundorf neundorf at kde.org
Mon Dec 13 18:57:39 GMT 2004


George wrote:
> > So IMHO preloading konqy slows the system on slow (ok, low memory)
> > systems down, on fast systems it is already fast. And the preloading adds
> > quite some opportunities for bugs (e.g. profiles)
> > So I'd vote for removing the preloading feature again.
>
> Why not just disable it in KControl?  

Well, that's what I did finally. No normal user would have known that 
disabling this option will free 50 MB of RAM. 

> In any case, is it possible to buy computers with less than 256MB RAM now?

On ebay, yes. 
I mean, it's a feature intended to make things faster, but on a low-memory box 
(well, 128 MB actually isn't that low) it makes the whole system slower. I 
think it is typically for slow systems that they come with not too much RAM, 
and for these systems optimizations are more important than for fast systems, 
but this feature makes slow systems even slower.


Allan wrote:
> Sounds like a kernel problem. Preloading shouldn't kill performance much, as 
> soon as konqueror is unused for a while, the kernel will swap it out and it 
> no longer has an impact on the remaining performance of the machine.

Well, as soon as it is swapped in again (because I want to visit a webpage) it 
has an impact again.

> If there is a way to detect memory presure we could ofcourse skip preload on 
> low memory machines, but I am not sure how to do it. 

> I disagree that it doesnt help on faster machines. It does.

Sure it does. But on fast machines it is even without preloading already fast.

Koos wrote:

> Wouldn't it make more sense to vote for fixing those memory leaks (and
> weren't those fixed for 3.3 already)?

I don't think these are memory leaks. It's just the memory used by the process 
which isn't given back to the OS until the process exits I think.

Bye
Alex
-- 
Work: alexander.neundorf at jenoptik.com - http://www.jenoptik-los.de
Home: neundorf at kde.org                - http://www.kde.org
      alex at neundorf.net               - http://www.neundorf.net




More information about the kfm-devel mailing list