RFC: Fedora Extras shipping ix86 optimized rpms?

Ralf Corsepius rc040203 at freenet.de
Wed Sep 1 09:23:01 UTC 2004


On Tue, 2004-08-31 at 10:36, Jakub Jelinek wrote:
> On Tue, Aug 31, 2004 at 10:30:33AM +0200, Arjan van de Ven wrote:
> > On Tue, 2004-08-31 at 10:21, Ralf Corsepius wrote:
> > > Hi,
> > > 
> > > Question: Shall Fedora Extras support ix86 optimized packages but
> > > i386-compiled packages rsp. under shall circumstances shall Fedora
> > > Extras support such packages?
> > > 
> > > Background: Some folks have started to add i686-built application
> > > packages in addition to i386-built packages to Fedora Extras, claiming
> > > these i686-built, "optimized packages" would result into much better
> > > performance of these packages ("up to factor 2").
> > 
> > those optimized packages aren't faster; at least I find it hard to
> > believe.... esp on p4 and athlon cpus where cmov is no gain again ;)
> 
> Well, SSE/SSE2 can help for graphic/video/audio applications.
If you say so, I don't have any reason for not believing you :-)

> But there .i686.rpm doesn't help you,
Can you explain?

It's the same approach RH applies to glibc and many 3rd party packagers
apply to their package.

My actual concern is less the technical side, but the "policy side" of
shipping "optimized packages" and its impact on "packaging"/"upgrading".
>  either the application
> selects whether to use SSE/SSE2 or not at runtime, or the packages can
> have separate sse2 and normal libs in one package:
> /usr/lib/libfoo.so.1
> /usr/lib/sse2/libfoo.so.1

i.e. a partial multilib implementation.

Packing-wise, this has several disadvantages.
1. You'd have to compile library packages twice.
2. Many packages contain both libraries and applications, so you'd have
to apply special measures to assure that applications still get
-march=i386 compiled. 
3. It would almost double the size of i386.rpms (These sse2 libs would
have to be part of i386.rpms) - Is it worth it?

However, I agree, it's a nice work-around suitable for libraries where
special optimizations can be proven to have a "significant/noticeable"
impact.

Finally, this doesn't cover applications - The initial sparc to ignite
this discussion was folks having entered "optimized applications" in FE
(cf. https://bugzilla.fedora.us/show_bug.cgi?id=2033)

Ralf






More information about the fedora-devel-list mailing list