rpmbuild - empty RPMS directory

[Dave Love]

arnaud gaboury <arnaud.gaboury at gmail.com> writes:

>> Why?  I ask as someone who supports R rpms on HPC systems and knows a
>> bit about relevant libraries and compilers.
>
> Please visit this blog[0] or many other talking about R-benchmark.R.

I don't think this is the place for a lecture on all that's wrong with
that; it's the sort of thing I try to educate users about.

>> If you want to subvert R's BLAS (and other slower versions) with
>> something largely optimal, look at the openblas-compat package under
>> <https://copr.fedoraproject.org/coprs/loveshack/livhpc/>.  [Yes, linear
>> algebra in Fedora is a shambles, with at least one failed attempt to
>> sort it out, but adding Intel stuff will only make things worse, and
>> that package finesses the issue fairly handily.]
>
> I agree than open-blas is a good soultion, but not as fast as the
> Intel BLAS libraries.

So are published, apparently-reproducible, results to the contrary by
OpenBLAS maintainers and others, wrong due to incompetence or
dishonesty?  Likewise for unpublished ones I've made in cases of
interest, and analysis of myths about Intel's code generation.

Anyhow, there's a demonstration that you don't need to rebuild Fedora
rpms to use a different BLAS.

-- 
Stand back I'm going to try SCIENCE  — XKCD