csnook at redhat.com
Sat Jan 12 22:21:41 UTC 2008
Jakub 'Livio' Rusinek wrote:
> I've used openSUSE (i586) with the same configuration and packages
> installed and eg. Firefox 3 was started in < 1 sec. I have no benchmarks
> but everything runs faster and is more responsible.
> Our optimalization does nothing compared to > i386 compilation.
False. On modern processors, with most userspace code, building for i586
generates code with instruction scheduling that is *slower* than i386
instructions tuned for i686. This is why Fedora builds most packages as i386,
tuned for i686.
i586 really only makes sense for things like the kernel, glibc, JVMs, and
multimedia/hpc libraries that use MMX/SSE instructions. In these cases, i686
will usually be even better, though it occasionally makes sense to build with
i586 instructions and tune for i686 if you want to have one package that'll work
acceptably on i586 and work very well on i686.
On a typical desktop system, you'll never notice the difference between i386,
i586, and i686 packages except with high-performance graphics. Your experience
with openSUSE is interesting, but it is not related to compiler optimizations.
> 2008/1/12, drago01 <drago01 at gmail.com <mailto:drago01 at gmail.com>>:
> 2008/1/12 Jakub 'Livio' Rusinek <jakub.rusinek at gmail.com
> <mailto:jakub.rusinek at gmail.com>>:
> > do we need to support legacy cpu's by i386 compilation?
> > i586 would make fedora faster even 3 times.
> > difference is noticeable.
> where are your benchmarks for the "3 times faster" claim?
> the i386 packages are already optimized for newer cpus (mtune vs. march)
> where it makes sense to have i686 versions there are some (kernel,
> fedora-devel-list mailing list
> fedora-devel-list at redhat.com <mailto:fedora-devel-list at redhat.com>
> Jakub 'Livio' Rusinek
More information about the devel