Re: ICE on linux-2.6 (2.6.18, official debs)
On Tue, Oct 24, 2006 at 12:31:41PM +0200, Roman Zippel wrote:
> > But I am very interested to hear about the distcc setup. Did I
> > understand it right, that every distcc needs one m68k box and a fast
> > cross-compiler box?
> Basically yes, but a single m68k machine usually doesn't keep a single
> fast machine busy, so multiple m68k machines could share one
> cross-compiler box.
Of course, and I don't think that such a shared crosscc box needs to be a
quad opteron. Any not-that-slow i386 box should be sufficient to speed up
m68k builds significantly...
In my distcc tests with m68ks, I got a speedup of roughly 100% when using
two m68ks (1x 060/1x 040) and compiling a kernel via 2 DSL links and VPN.
So, I guess using hosts on the local network won't be slower... ;)
> > So even the slower macs could become really useful as
> > long as they have decent network hardware?
> Yes, but we maybe have to be a bit careful where to queue packages, it
> only speeds up c/c++ compilations. OTOH it could free up the big machines
> for the packages, which still need this, e.g. java, objc or haskell
> packages. Cross compiling objc is possible, but it's not packaged yet, ghc
> unfortunately uses -x a lot and also produces large generated files, so
> it's a bit of a PITA to compile.
It's still a question how to integrate a distcc/crosscc into our existing
buildd environment. I doubt that we want to do this all by manual builds...
;)
--
Ciao... // Fon: 0381-2744150
Ingo \X/ SIP: 2744150@sipgate.de
gpg pubkey: http://www.juergensmann.de/ij/public_key.asc
Reply to: