[Numpy-discussion] [ANN]New numpy, scipy and atlas rpms for FC 5, 6 and 7 and openSUSE (with 64 bits arch support)

David Cournapeau david at ar.media.kyoto-u.ac.jp
Thu Jun 28 23:15:24 EDT 2007


Eike Welk wrote:
> On Sunday 24 June 2007 13:05, David Cournapeau wrote:
>> Hi there,
>>
>>     After quite some pain, I finally managed to build a LAPACK +
>> ATLAS rpm useful for  numpy and scipy. Read the following if you
>
> ------- snip --------------------------------------------------
>
>> and lapack. I would like to hear people complains. If people want
>> other distributions supported by the opensuse build system (such as
>> mandriva), I would like to hear it too.
>
> I tried your repository 
> (http://software.opensuse.org/download/home:/ashigabou/openSUSE_10.2/) 
> with two machines running openSUSE 10.2: 
> 1. AMD Athlon desktop
> 2. Pentium-M laptop. 
>
> The repository works with Yast (installation program). 
> The prebuilt packages work on both machines. They especially work with 
> matplotlib from the http://repos.opensuse.org/science/ repository. (I 
> didn't try timers and testers.)
>
> Building Atlas succeeds on the Pentium-M (2). On the Athlon (1) the 
> sanity checks fail. 
>
> The resulting Atlas RPM is missing two links:
> /usr/lib/atlas/sse2/libblas.so.3   -> /usr/lib/atlas/sse2/libblas.so.3.0
> /usr/lib/atlas/sse2/liblapack.so.3 -> /usr/lib/atlas/sse2/liblapack.so.3.0
The atlas rpm is still really rough on the edges. Ideally, you should 
not need LD_LIBRARY_PATH, and I should update the loader (the software 
which looks for shared libraries when launchdng a program) cache, but if 
I make a mistake at this point, this has the potential to screw up the 
whole machine....

Could you give me more information on the AMD failure ? Are you using 64 
bits mode ?

>
> When I try your one line examples, I see a nine times speedup with 
> Atlas.
>
> Thank you for your efforts! You provide an easy way to install Atlas 
> on Suse for the first time.
>
> Regards,
> Eike.
>
> PS.: 
> I still think you should contribute to the 
> http://repos.opensuse.org/science/ repository. Then this quite 
> comprehensive repository would get decent Blas and Atlas too.
This is planned, and I already took contact with them a few weeks ago, 
but I didn't have time to do it properly yet.

David



More information about the NumPy-Discussion mailing list