Hi Danilo,
Likewise, very interesting, I'll certainly be there!
Besides the things mentioned by Alan below, to have a truly reproducible
setup, EasyBuild should be combined with Nix (see
https://nixos.org/nix/), in order to be in full control of the build
environment.
Here's a very interesting paper on the use of Nix in an HPC context:
http://arxiv.org/pdf/1506.02822.pdf .
They discuss some of the shortcomings EasyBuild has w.r.t.
reproducibility, and they do a good job at it.
I feel they're not fully grasping the power of EasyBuild, but I can
forgive them, I really like the paper.
I hope to get them to submit a talk to the HPC & Big Data devroom at
FOSDEM'16 I'm co-organizing (see http://hpc-bigdata-fosdem16.github.io).
And since we're on the topic of BoF sessions at SC15, here's a link to
our own "Getting Scientific Software Installed" BoF (Tue Nov 17th,
12.15pm): http://sc15.supercomputing.org/schedule/event_detail?evid=bof113 .
Tell your colleagues to join us there, and bring up the topic of
reproducible builds.
I'll be happy to let them advertise their BoF which is shortly after ours.
regards,
Kenneth
On 10/11/15 19:07, Alan O'Cais wrote:
Hey Danilo,
I won't be at SC but this topic is something that interests me. As
EasyBuild develops, some features make reproducability better (like
becoming as independent from the system compiler as possible) while
others are introducing potential unknowns (--include-easyblocks,
--include-toolchains, --minimal-toolchains). I've already been
thinking about making a PR that attempts to improve reproducability by
including the toolchain and easyblock classes (*that were actually
used* when producing the build) in the easybuild directory of the
installation along with the easyconfig of the installation that
already exists there. I think with some tweaking and a few more
features this could be very useful.
Best,
Alan
On 10 November 2015 at 17:24, Danilo Guerrera
<[email protected] <mailto:[email protected]>> wrote:
Hello everybody,
some of you will be in Austin during the next week. I exploit this
mailing list to sponsor a BOF organised by the group, High
Performance and Web Computing of the University of Basel, where I
am a PhD candidate.
We are developing a tool targeting reproducibility in
computational sciences, allowing collaboration between users.
During the last SC I discovered EasyBuild and lately decided to
rely on it for getting “reproducible builds”. That’s why we are
integrating it as a basis for our tool “PROVA!”.
I won’t be at SC this year but my supervisor, prof. H. Burkhart,
is organising the BOF in collaboration with prof. G.Wellein from
FAU-Erlangen. It will be a good occasion for talking of the
difficulties for reproducible research.
Everyone is more than welcome to join the session:
http://sc15.supercomputing.org/schedule/event_detail?evid=bof145
Regards,
Danilo
Departement Mathematik & Informatik
High Performance and Web Computing Group
Universität Basel
Spiegelgasse 1
CH-4051 Basel Switzerland
email: [email protected] <mailto:[email protected]>
Phone:+41 (0)61 267 14 73 <tel:%2B41%20%280%2961%20267%2014%2073>
--
Dr. Alan O'Cais
Application Support
Juelich Supercomputing Centre
Forschungszentrum Juelich GmbH
52425 Juelich, Germany
Phone: +49 2461 61 5213
Fax: +49 2461 61 6656
E-mail: [email protected] <mailto:[email protected]>
WWW: http://www.fz-juelich.de/ias/jsc/EN
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich
Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher
Geschaeftsfuehrung: Prof. Dr.-Ing. Wolfgang Marquardt (Vorsitzender),
Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
Prof. Dr. Sebastian M. Schmidt
------------------------------------------------------------------------------------------------
------------------------------------------------------------------------------------------------
