[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: SPARC Status (Was Re: Secondary Architecture Status?)



On Tuesday 12 May 2009 10:06:22 Peter Robinson wrote:
> >> > I would really like to see a proliferation of secondary arches in
> >> > Fedora, but I don't think 'workstation' is a viable usage model for
> >> > them to get started.  Most will have to focus on the type of hardware
> >> > that actually sells for that arch, and yes I realize that can be at
> >> > odds with some of the directions Fedora is going.
> >>
> >> I think there are most likely two candidates for a secondary arch
> >> other than PPC. The first is sparc where from the server point of view
> >> where its probably about as prolific as PPC in that regard. The second
> >> would be arm but that is more from the NetBook/MID/Phone/STB
> >> perspective where there are quite a few devices in the 500Mhz-1Ghz
> >> range with 256-512Mb RAM. With the OLPC X0-1 we've proven that Fedora
> >> can run relatively well on that sort of spec, they also tend to be
> >> relatively cheap.
> >
> > Once we switch our 32-bit x86 packages from i586 to i686 for F12
> > (hopefully w/sse2 enabled as well), then i586 becomes a very doable
> > secondary arch as well, if enough people are interested in keeping
> > their clunkers running the latest-n-greatest stuff...
> 
> What about the million odd (not sure of the actual deployment numbers)
> OLPC XO clunkers out there that fall into this category?

The fact the next olpc release is to be based on f11 is in fact one of
the reasons f11 went i386->i586 for 32-bit x86 binaries... That camp
might take a great deal of interest in an i586 secondary arch. :)

> I though the
> whole difference between i586 and i686 was cmov and that wasn't
> considered a great performance boost.

Well, that's one part of it. With luck, i686 will actually be cmov
plus sse2, which is a fairly decent win.

Also, Ulrich and Jakub (iirc) were rather leery of running as i586,
as the i586-codepaths aren't as well exercised or optimized as
i686-specific ones (I don't have any specific examples though).

-- 
Jarod Wilson
jarod redhat com


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]