KDE-Forum.org

Quoted

Original von seb

Found this useful article on kde-wiki.

http://wiki.kde.org/tiki-index.php?page=Performance+Tips

If I understood the discussion on the mailing lists correctly this is a draft version that will move somewhere else when it's finished. So we'll have to check this link from time to time and then adapt it.

Is this the right place to talk about optimizing KDE in terms of GCC compiler options? I use konstruct to build KDE myself. The default optimization flags are "-O2 -pipe". I am wondering if I can speed up KDE in general by using a more aggressive flag such as O3 or CPU-specific flags such as march=pentium4? "O3" will probably increase the code size, but I have plenty of hard disk space.

TIA,
Dai

compiling with gcc 3.4.4 or gcc 4.0.1 improves kde performance.

If you are not worried about 'crash reports'/debug info, and do not use IPV6, I found these two environment $VARS speed things up a tad:

KDE_DEBUG=0
KDE_NO_IPV6=1

BTW, 3.5.1 is _perfect_!! All the little issues I had/bug reported got fixed!

Nick

daihard, I think it's off-topic, but still like to give some answers:

Quoted

Originally posted by daihard
Is this the right place to talk about optimizing KDE in terms of GCC compiler options?

I'm not sure, but do you think it really improves the code a lot? It's still the same code... and gcc is still one of the most inefficient compilers.

And consider the fact your distribution already compiled KDE with -O2, and likely -mcpu=i686. Your distribution likely took care of "unused direct dependencies" too, in constrast with the default behavour of ./configure scripts. Try the commands:
At SuSE here binaries are already linked against the bare minimum. When you compile things yourself you'll need to fix that too.

Quoted

The default optimization flags are "-O2 -pipe". I am wondering if I can speed up KDE in general by using a more aggressive flag such as O3 or CPU-specific flags such as march=pentium4? "O3" will probably increase the code size, but I have plenty of hard disk space.

Huge binaries also increase the loading time of the application. So what's more to gain here? A desktop application is idle 99% of the time. The times you need speed is at start-up and with some batch-processing operations.

It may sound a bit depressing, but I like to be realistic about these topics. Things like Qt4, a better compiler, better C++ support in gcc and better usage of the harddrive-IO improve a lot more then some tweaking with flags.

I do not know more about it also i do not know how it will be compiler more effectivelly.so Please tell me Abount it anyone who is so good in this field.