compilation architecture

[Michael Schwendt]

On Wed, 16 Jan 2008 16:56:58 +0100, Jakub 'Livio' Rusinek wrote:

> > On Wed, 16 Jan 2008 14:09:13 +0100, Jakub 'Livio' Rusinek wrote:
> >
> > > Ok, I've installed openSUSE and configured it like my Fedora.

I went through the pains of installing openSUSE 10.3 for GNOME. Except for
the installation settings, which I adjusted only minimally, I used the
defaults and I did not reconfigure anything after booting the first time.

I explicitly did not enable the non-free repository.

First thing I noticed: harddisk access in openSUSE happens at half the
speed of Fedora. hdparm -t confirms it whereas hdparm -T results for
cache reads are about the same for both systems.

This difference in disk access had a bad effect on some data processing
speed tests. For huge files, openSUSE was about 33% slower in overall
execution time, and I almost stopped at that point.

For tests with smaller files, openSUSE computed approx. 3% faster than
Fedora 8 (this is on AMD Athlon). Sometimes 2%, sometimes a bit more than
3%, but I didn't spend the time to take more than 1 or 2 samples each. And
it was unconvincing enough to even consider comparing special computations
of e.g. ASM-optimised A/V codecs.

The openSUSE GNOME Desktop does not feel any snappier to me. Not when
starting applications and not when opening xterm either. Though, I've
noticed that when dragging xterm from the menu onto the desktop, an
invalid launcher file is created and gives an error. ;)

Firefox (it's 2.0.0.6 in openSUSE and 2.0.0.10 in F8) in Fedora 8 starts
and exits from within a terminal in roughly 4.3 seconds (reproducibly)
when quickly pressing Ctrl+W as soon as the window appears. That should
mimic your own test procedure. In openSUSE it did that in approx. 6
seconds, and possibly due to the disk access issues I had difficulties in
trying to start and close it faster than in 5-6 seconds.