Re: atlas, lapack, blas
On Fri, Apr 09, 2010 at 06:30:22PM +0200, Sylvestre Ledru wrote:
> > Likewise no attempt at configuring ld.so for sane defaults on the
> > running cpu are attempted.
>
> I didn't know it was possible before someone pointing this out on my
> blog.
>
> > ldd /usr/bin/octave-3.2.4|grep blas
> > libblas.so.3gf => /usr/lib/libblas.so.3gf (0xb57f9000)
> >
> > 2) The tester and performance binaries are gone.
>
> My bad. I will put them back in a future upload.
Hi, I'm wondering what's happening with the amd64sse3 versions of the
atlas libraries?
an apt-get dist-upgrade wants to remove the amd64sse3 packages due
to the upgrade of libatlas3gf-base:
The following packages will be REMOVED:
libatlas-amd64sse3-dev (3.8.3-24)
libatlas3gf-amd64sse3 (3.8.3-24)
[...]
The following packages will be upgraded:
[...]
libatlas-dev (3.8.3-24 => 3.8.3-27)
libatlas3gf-base (3.8.3-24 => 3.8.3-27)
are these packages coming back in a future version, or does the
libatlas3gf-base package now include libs optimised for opteron?
i've been holding off any upgrades for a few months now, waiting to see
what's happening. not complaining, just wanting to get an understanding
of what the plans are for atlas.
(and if you need access to an opteron machine to compile opteron
versions, i can arrange an account for you on an 8-core machine.)
I built a 20-node cluster (for computational chemistry) at work earlier
this year using sunfire X2200 M2 quad-core opteron machines, so getting
the best performance possible out of them would be a Good Thing.
craig
ps: i want to say a big Thank You to the debian-science teams. I've
been a debian user and DD since 1994 (so i'm already biased towards
debian :), but debian's excellent scientific software packaging has
really made it an easy best choice for scientific workstations where i
work, and the slurm-llnl and openmpi and other cluster-related packages
made it a great choice for the cluster too.
--
craig sanders <cas@taz.net.au>
Reply to: