KDE4 slow

[Michael Pyne]

On Friday 26 June 2009 11:06:07 Baas, Kevin wrote:
> >if it were that simple. in the case of the "plasma is slow"-"it was
> > actually kio_http" case, no amount of timestamps in plasma would have
> > identified the reported problem.
>
> not true.  if kio_http was serially connected to plasma, it would have been
> inbetween two timestamps, telling u where the problem was.  

Except that unless you knew to look for kio_http then you wouldn't necessarily 
have been receiving any debugging output from kio_http.  At best your 
profiling data would have clustered around Plasma methods blocking on 
kio_http.

> if it was a
> separate process/thread spawned asynchronously, then your timestamps would
> tell you that the problem was not in plasma.  (and, in fact, plasma would
> not be slow.)

Well, kioslaves always run in a different process so this is actually what was 
happening.  You're right that the timestamps would show nothing in Plasma, in 
fact that is Aaron's point.  The problem is that, again, unless you knew to 
look for kio_http specifically, (which is separate process entirely in this 
case) then it would be very hard to tell why Plasma is slow.

And in fact, it is certainly possible for Plasma to be "slow" even while 
taking very little CPU if it were waiting for output from kio_http.  Operating 
systems nowadays are designed to ensure that processes waiting for I/O don't 
use up hardly any CPU time, so the timestamps wouldn't show CPU offenders in 
Plasma because there wouldn't be any.

> welcome to the world of debugging.

Yeah, we've been here for quite a few years, welcome yourself. ;)

Regards,
 - Michael Pyne
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://mail.kde.org/pipermail/kde-quality/attachments/20090626/86b32abd/attachment.sig