[PyKDE] PyKDE concatenation and gcc4.0.3 [was: New PyKDE snapshot - help needed]

[Stephan Hermann]

Hi Andreas,

On Friday 20 January 2006 21:24, Andreas Pakulat wrote:
> On 20.01.06 20:53:13, Stephan Hermann wrote:
> > Hi Jim,
> >
> > On Friday 20 January 2006 20:03, Jim Bublitz wrote:
> > > On an 800MHz machine with a 100MHz front side bus, the concatenated
> > > version using gcc 3 takes about 45 minutes, the non-concatenated
> > > version with gcc 4 takes about 75 minutes, and (as I recall) the
> > > non-concatenated version with gcc 3 takes close to 3 hours. So gcc 4 is
> > > still much faster, even without concatenation. Thanks for the info on
> > > 4.0.3 - I'll modify configure.py to take that into account
> > > automatically.
> >
> > on my amd64 sempron 1.6GHz with 512 Megs, the concatenated versions takes
> > more then 3 hours. Because it uses all available RAM and a huge ammount
> > of swap space. The normal file by file compile takes only 1 hour.
>
> Wow, that's extreme. I compiled PyKDE on a Centrino 1.4 GHz in less then
> 30 Minutes having only 512 Megs Ram (and 700 Swap), while actually
> running KDE, including amarok (which is for a music player quite
> cpu-hungry) and writing some mails (in a xfce4-terminal, which is also
> quite cpu-intensive). As I said, the system slowed down really good but
> there were still at least 30 Megs Ram unused and I don't think that any
> swap got used for that.
>
> Oh that's the concatenated version of PyKDE I'm talking about, the
> non-concat version is taking about double or tripple the time, but uses
> much less memory.

Yes, I have to run more tests about this issue. But I had the same issues when 
I used in the past a 1.6GHz Pentium M with 512 Megs (HP nc6000 laptop). So, 
it can be, that all this is a regression of something in the toolchain, or 
the build environment (tested both builds in pbuilder on ubuntu) is doing 
something else, then compiling the source directly on the system. But I 
compiled the i386 test packages in an i386 chroot on this amd64 (with -i 
option) and it compiled much faster then in the pbuilder environment (without 
-i). I'll test it again, until I find a reasonable answer what causes all 
these strange result.

Regards,

\sh