Le mardi 22 juin 2010 à 11:18 +0200, Samuel Thibault a écrit : > Sylvestre Ledru, le Tue 22 Jun 2010 10:53:40 +0200, a écrit : > > Le mardi 22 juin 2010 à 10:36 +0200, Samuel Thibault a écrit : > > > Sylvestre Ledru, le Tue 22 Jun 2010 10:15:35 +0200, a écrit : > > > > Le mardi 22 juin 2010 à 10:01 +0200, Samuel Thibault a écrit : > > > > > Sylvestre Ledru, le Tue 22 Jun 2010 09:20:18 +0200, a écrit : > > > > > > Le dimanche 20 juin 2010 à 00:55 +0200, Kurt Roeckx a écrit : > > > > > > > On Wed, Jun 16, 2010 at 11:22:41PM +0200, Sylvestre Ledru wrote: > > > > > > > > Le samedi 12 juin 2010 à 00:25 +0200, Samuel Thibault a écrit : > > > > > > > > > Sylvestre Ledru, le Sat 12 Jun 2010 00:01:25 +0200, a écrit : > > > > > > > > > > Does anyone see a better way of handling this specific > > > > > > > > > > problem ? > > > > > > > > > > > > > > > > > > Can't atlas just build the binaries and not try to run then? > > > > > > > > No because Atlas is building many time the same sources to see > > > > > > > > which > > > > > > > > optimisation performs best... > > > > > > > > > > > > > > Does that mean it's optimizing for the hosts that build it? > > > > > > It is indeed the way atlas is working... > > > > > > The best way to achieve maximum performances is to build locally > > > > > > atlas. > > > > > > No doubt about that. > > > > > > > > > > Then can't there be a flag to just select a > > > > > not-so-optimised-but-generic > > > > > version? > > > > This task is done by the base package of atlas. > > > > > > I mean, including sse2, sse3 etc. generic versions too. > > Sorry but I don't understand your suggestion ?! > > > > When Atlas is built locally, it will detected automatically what are the > > extensions available on the CPU.. > > I understand that. What I mean is that a buildd shouldn't _need_ to have > sse3 capabilities to be able to build the sse3 package. I agree on the principle (and I really wished it was the case) but it is not the way Atlas build system is working ...
Sylvestre -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected] Archive: http://lists.debian.org/[email protected]
