[Q-e-developers] [Pw_forum] QE installation on standard HPC machines.

[Filippo Spiga]

Ciao Emine,

you idea is not bad at all. Having a set of different reference outputs will surely help users to understand and maybe detect what went wrong during the configure/compile process on their machines. Especially for Tier-0 or Tier-1 machines in Europe (Tier-0 and Tier-1) where there is such a wide diversity of systems and environments. Lot of people sometimes give up or waste a lot of time for something that has to be easy and fast.

I am afraid, the real problem/challenge is the environment. Every system has its own environment and set of modules and specific paths where libraries and compiler are installed. Sometimes when it changes it is a real pain. In any case, starting to collect feedback will help to improve the configure and reduce the necessity to edit manually the make.sys (a very dangerous thing if someone is not expert a bit…)

Carlo works mainly on IBM systems and he has plenty of examples about how compile & run on BlueGene/P or BlueGene/Q. I have an access to several CRAY XT/XE/XK systems so I can work on it (if you have a problem using CSCS clusters send me an email directly).  We need to agree about a file format/name convention before put anything in the repository. 

Paolo, Layla… what do you think about this?

Cheers,
F.

On Aug 7, 2012, at 10:38 PM, Emine Kucukbenli <kucukben at sissa.it> wrote:
> ----- Forwarded message from Emine Kucukbenli <kucukben at sissa.it> -----
>    Date: Wed, 25 Jul 2012 16:15:48 +0200
>    From: Emine Kucukbenli <kucukben at sissa.it>
> Reply-To: PWSCF Forum <pw_forum at pwscf.org>
> Subject: [Pw_forum] QE installation on standard HPC machines.
>      To: pw_forum at pwscf.org
> 
> Hi all,
> 
> I have decided to gather working make.sys files, configurations for all the standard HPC machines that are commonly used by the community.
> 
> My purpose is to build a reference database for installation procedures; organize wrt machine/compiler/version/configure options and then share it; to make life easier for all; users, mailing list responders and hpc admins.
> 
> There is a similar thing in /install directory but it doesnt seem very up to date, and i think a division with machine would really be helpful for starters.
> 
> So I would be glad if you can help me construct this reference database
> by sending your make.sys and config.log files which you already have;
> and further details if necessary, such as the version, list of loaded modules, test status etc.
> I would also like to acknowledge the contributors so that people using the same machine can have an opportunity to get to know each other.
> 
> Ok, some examples from the complete list is going to look like the following:
> 
> CINECA_fermi_intel_hybrid_make.sys and CINECA_fermi_intel_hybrid-config.log
> CSCS_rosa_pgi_mpionly_make.sys CSCS_rosa_pgi_mpionly_config.log
> etc.
> just to give an idea.
> 
> I have also added my CEA_CURIE make/log files here as an example.(on CURIE QEv5.0 work out of the box atm, so it is not that useful but still..)
> 
> If you have any questions, or if you think this already exists, or is unnecessary,or bad for IT security etc, please let me know.
> 
> Looking forward to hearing from you!
> Thanks!
> 
> emine kucukbenli
> former phd student, sissa, italy
> address change to: postdoc, epfl, switzerland.
> 
> ----- End forwarded message -----

--
Mr. Filippo SPIGA, M.Sc., Ph.D. Candidate 
CADMOS - Chair of Numerical Algorithms and HPC (ANCHP)
École Polytechnique Fédérale de Lausanne (EPFL)
http://anchp.epfl.ch ~ http://filippospiga.me ~ skype: filippo.spiga

«Nobody will drive us out of Cantor's paradise.» ~ David Hilbert