Hello,
I just noticed the gnu-hpc packages in Tumbleweed recently, and I was
wondering if there is an explanation of how they work somewhere? I had
a look for some information or discussion, but I've not been able to
find anything so far.
I gather they are linked somehow to environment modules, but it's not
clear to me how exactly. For instance, "petsc-devel" installs a module
for PETSc, but "hdf5-devel" does not install one for HDF5. Are there
certain packages we're supposed to install to get this functionality?
Also, are they supposed to play nice with pkg-config? petsc-devel does
install PETSc.pc, but loading the module doesn't update
$PKG_CONFIG_PATH to be able to find it.
Cheers,
Peter
--
Dr Peter Hill
Associate Research Software Engineer
York Plasma Institute, Department of Physics
University of York
Heslington
York, YO10 5DD Tel +44 (0)1904 324902
Physics Coding Club: https://physicscodingclub.github.io
Disclaimer: http://www.york.ac.uk/docs/disclaimer/email.htm
--
To unsubscribe, e-mail: opensuse-science+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-science+owner(a)opensuse.org
Todd Rme writes:
> With the new multibuild approach, I am wondering if it might be time
> to revisit MKL support in openSUSE. It is not open-source, which may
> be a deal-killer in general. But it is freely available to open-source
> projects, and it can improve computational performance considerably at
> least on intel CPUs. It couldn't serve as a replacement for lapack or
> openblas, which is part of why support for it was rejected previously,
> but with multibuild it may be possible to support it alongside
> existing implementations.
If you want to build packages against MKL in OBS, it would have to be
packaged and available in OBS if you want to build dependent projects.
Generally it is possible to package binary blobs, however, I'm not sure
if the license would allow this. And even if it did, it is sorta-a
frowned upon.
I suppose it's possible to add it to the spec file as yet another flavor.
To keep things manageable, it would make sense to build this using
environment modules.
the RPM macros I made for this currently only support building against
different flavors of MPI libraries, however.
I was briefly contemplating about generalizing this concept so that
building against different flavors of Lapack and BLAS could be used
as well, however, it was complex enough as it was working against the
recursion limit of 16 in the RPM macro parser which required some
tricks.
Apparently the next version of RPM will have this lifted to 64 (yeah!)
so this could be revisited then.
Cheers,
Egbert.
--
Egbert Eich (Res. & Dev.) SUSE LINUX GmbH
SUSE Labs - Project Manager HPC
Tel: +49 911-740 53 0 http://www.suse.com
-----------------------------------------------------------------
SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
Maxfeldstraße 5, D90409 Nürnberg, Germany--
To unsubscribe, e-mail: opensuse-science+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-science+owner(a)opensuse.org
With the new multibuild approach, I am wondering if it might be time
to revisit MKL support in openSUSE. It is not open-source, which may
be a deal-killer in general. But it is freely available to open-source
projects, and it can improve computational performance considerably at
least on intel CPUs. It couldn't serve as a replacement for lapack or
openblas, which is part of why support for it was rejected previously,
but with multibuild it may be possible to support it alongside
existing implementations.
--
To unsubscribe, e-mail: opensuse-science+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-science+owner(a)opensuse.org