comments on KDE performance tips

Allan Sandfeld Jensen kde-optimize@mail.kde.org
Tue, 14 Jan 2003 00:23:41 +0100


On Monday 13 January 2003 14:56, Alexander Kellett wrote:
> On Mon, Jan 13, 2003 at 03:13:19PM +0100, Dirk Mueller wrote:
> > fast-malloc is default, and --enable-final shouldn't give you significant
> > speed advantage when actually running the build. --disable-debug does so,
> > though.
>
> i've not done any real-world profiling but doesn't enable-final speed
> up not only compiles but also builds due to the fact that gcc can inline
> more relevant code that otherwise?, probably decreases object size also.
> but if its enough difference to be seen, i'm not sure.
>
No, it takes _alot_ longer to compile, but might give minor speed-ups. That's 
why it is default off.

The Intel compiler has much better intraprocedure call optimizations and 
doesnt even need the sources files cat'ed to do them.

Once libtool in KDE gets upgraded to the most recent libtool-cvs, my patch for 
Intel C++ supported should be in. My own version of KDE 3.1 compiled with icc 
works fine except for artsd that has some sort of assertion crash. Although I 
am not sure if the Intel Compiler really makes KDE faster (hard to 
benchmark).

`Allan