Re: kde cygwin patches

> > with gcc 3.3.3 (3.4.3 was a bad idea, although the
> > built code ran MUCH faster than 3.3.3) so far. 
> You have written about dynamic_cast problems. Are
> there more problems using 
> kde 3.4 ? Is this problems could be handled this
> seems to be very 
> interesting, because I recognized performance
> problems for example with 
> umbrello. The most time when importing classes is
> gone in the dynamic_cast 
> related functions (about 60%), so I assume, that
> this cast operations are 
> very bad implemented in gcc 3.3.3. 

It seems like its not an issue with the gcc version.
When I compiled with gcc-3.4.3, I had taken out
certain dynamic_cast from the code and replaced them
with static_cast, to avoid certain crashes. It was
probably because I needed some cygwin patches for
dynamic_casts to work properly. So, the code ran MUCH
faster when those dynamic_casts were NOT around in the
code.

so your analysis that dynamic_casts are where most of
the slowdown for  KDE is, is true. And from my
experience of seeing the KDE startup, its a huge
slowdown.

After I took the dynamic cast patch from 3.4.1
package, built the gcc 3.4.3 and re-compiled kde, it
became as slow as with gcc-3.3.3.

Is there anyone looking at the possibility that
whether those many dynamic_casts are really required
in the KDE code? That many dynamic casts usually means
bad design.

particularly, if it can mean speed up for even
non-cygwin platforms, it is worth the investigation.

-Sunil


		
__________________________________ 
Do you Yahoo!? 
Yahoo! Small Business - Try our new Resources site
http://smallbusiness.yahoo.com/resources/