[Kst] Performance

[George Staikos]

On Thursday 16 December 2004 13:55, Barth Netterfield wrote:
> On December 16, 2004 01:06 pm, George Staikos wrote:
> > On Thursday 16 December 2004 13:01, Barth Netterfield wrote:
> > > > > -Make sure things aren't waiting for things that they don't need to
> > > > > be waiting for...
> > > >
> > > >     Include X11 calls, which is what QCursor::pos() does.  It's a
> > > > roundtrip...
> > >
> > > How long on a local machine?  Over the web, obviously, this is open
> > > ended... 1 per mouse I assume...
> >
> >   10-35ms per call, local client.  I have a patch that removes most of
> > these but it's not done yet.  (Still buggy)
>
> Ahhh... so we were doing lots of them unneededly.  Are you just cacheing
> it? This could be a huge speedup for multi-plot situations (as you have
> been trying to say.)

  Exactly! :-)  Yes, we were actually already caching it, but there were some 
calls missing to keep things in sync.  I have a patch that is at least "close 
enough" to working to commit.  There might be some small quirks, but I 
haven't seen anything major yet.  I'm working from the laptop right now and 
it's too loaded to give me good numbers but if you want to test, just check 
to see what the difference is in paint() when there is no draw.  It also 
affects the case where there is a draw (ie: dirty), but not as big 
percentage.

   I noticed that the tie boxes flicker a whole lot here.  We really need to 
sort this out somehow.  Our painting for data mode and tie boxes is really 
sub-optimal.

-- 
George Staikos
KDE Developer				http://www.kde.org/
Staikos Computing Services Inc.		http://www.staikos.net/