[Bug 659896] Review Request: cp2k - Molecular simulations software

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=659896

Jussi Lehtola <jussi.lehtola at iki.fi> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
                 CC|                            |jussi.lehtola at iki.fi
         AssignedTo|nobody at fedoraproject.org    |jussi.lehtola at iki.fi
               Flag|                            |fedora-review?

--- Comment #1 from Jussi Lehtola <jussi.lehtola at iki.fi> 2010-12-04 18:13:20 EST ---
Hmm, I've meant to package this for some time, although I haven't been using it
myself for the time being. In the past one of the problems was that there were
no stable releases. This seems to have changed somewhat as they now have a
stable branch. However, one might have to be careful in picking the snapshots.

I would put the OpenMP parallelized version in the main package, myself, as it
can be ran in single thread mode anyhow.

Another thing is that OpenMP parallellism is usually not as efficient as MPI
parallellism on NUMA architectures due to the used memory not being optimally
located. Getting the MPI parallellized version to build, however, may not be a
trivial task, e.g., as it needs scalapack.

*Please* don't build against reference LAPACK, you'll get an order more
throughput by building against ATLAS. Replace BR: lapack-devel and BR:
blas-devel with BR: atlas-devel. You'll need to pass "-L%{_libdir}/atlas" as
LDFLAGS and "-lf77blas -latlas" as the BLAS library and "-llapack" as the
LAPACK library.


I'll go through the review during the weekend.

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.