Skip to content
2017-10-12.rst 6.12 KiB
Newer Older
Victor's avatar
Victor committed
TelCon 12 of October
====================

Apologies:
----------
 - Charles Moulinec (STFC)
 - Dimitris Dellis (GRNET)
 - Mariusz Uchronski (WCNS/PSNC)

Present:
--------
 - Victor Cameo Ponz (CINES)
 - Luigi Iapichino (LRZ)
 - Remi Lacroix (IDRIS)
 - Andrew Emerson (CINECA)
 - Philippe Segers (GENCI)
 - Jacob Finkenrath (CyI)

Minutes of meeting:
-------------------

Update on machine availability & login procedures
*************************************************

 - GPU prototype is up and running in Cineca. There is still some imprecision
   concerning access for 4IP-extension: it might delay again for 10 days, but
   we are currently negociating that point so it's faster. The login procedure
   will not go through the userdb but A. Emerson will gather access requests
   (name, email address and affiliation required). Action for VCP: send the BCOs
   list to Andrew. Namely:

    - EMERSON Andrew - CINECA
    - PROEME Arno - EPCC
    - DE LA CRUZ Raul - BSC
    - FINKENRATH Jacob - CyI
    - UCHRONSKI Mariusz - WCNS/PSNC
    - DELLIS Dimitris -  - GRNET
    - MOULINEC Charles - STFC
    - CODREANU Valeriu - SURFsara
    - CAMEO PONZ Victor - CINES

   Since GADGET and GPAW are not supposed to be run on GPU, BCOs have not been
   included.
 - No news from FPGA prototype staff.


Basics for energy mesurement on BULL/Atos KNL machine
*****************************************************

BEO is fully functionnal, and now HDEEVIZ is starting to work if you want to try
it. There were a workshop at CINES last week where trainees used them successfully.
More documentation on ``/opt/software/frioul/documentation/``. Contact the support
team at CINES if you have any requests (svp@cines.fr).


Figures that will be included in the final deliverable (performances & energy related)
**************************************************************************************

I'm working on a template that all of you will fill with performances and energy
metrics. Comparison will be base on full nodes so that we can compare results regarding
theoretical peak performances, memory characteristics and so on. At least on set of
metrics will be required by test case by machine, and if you have more (ie if you have
scalability) it's welcome obviously.
Also some comments on performance will be required.


General 4IP-extension information
*********************************

I just updated the milestone so it mention the project of INRIA on the KNL machines.
This is a side project focusing on one PRACE code and it's energetic optimisation.


Questions, concerns and report from BCOs
****************************************
Everything about login opening, deadlines, expected work during the next period, others:

   +------------------------+-----------------------------------------------------------+
   |                        |                                                           |
   |   Code name            +   4IP-extension BCO                                       +
   |                        |                                                           |
   +========================+===========================================================+
   | ALYA                   | No feedback                                               |
   +------------------------+-----------------------------------------------------------+
   | Code_Saturne           | Lot of work lately, will begin 4IP-extension related work |
   |                        | not before 1 week from now                                |
   +------------------------+-----------------------------------------------------------+
   | CP2K                   | No feedback                                               |
   +------------------------+-----------------------------------------------------------+
   | GADGET                 |  Concerns about UEABS version vs current one: contact     |
   |                        |  Walter Lieon on this subjet. Currently compilation some  |
   |                        |  problem, more updated version of the code might help     |
   +------------------------+-----------------------------------------------------------+
   | GPAW                   |  No feedback                                              |
   +------------------------+-----------------------------------------------------------+
   | GROMACS                | Runs almost finished. Some problem at the end of the jobs |
   +------------------------+ might make it difficult to get the energy back. BEO usage +
   | NAMD                   | WIP, send the SSH key but no feedback yet                 |
   +------------------------+-----------------------------------------------------------+
   | NEMO                   |  No feedback                                              |
   +------------------------+-----------------------------------------------------------+
   | PFARM                  |  compiled and executed PFARM code on KNL machine.         |
   +------------------------+-----------------------------------------------------------+
   | QCD                    |  No updates, waiting to get some time to work on QCD      |
   +------------------------+-----------------------------------------------------------+
   | Quantum Espresso       | Compil OK knl. Few runs. Work fine. Currently running     |
   |                        | flat mode. Would certainly had some perfomance gain with  |
   |                        | cache mode                                                |
   +------------------------+-----------------------------------------------------------+
   | SHOC                   |  No feedback                                              |
   +------------------------+-----------------------------------------------------------+
   | Specfem3D_Globe        | Compiled and began to run. But facing I/O issues so       |
   |                        | no metrics for now. Action: contact France Boillod at     |
   |                        | Atos and Bertrand Cirou at CINES                          |
   +------------------------+-----------------------------------------------------------+

AoB - Date of next meeting
****************************
 - 2 of November, 11:00 CEST, primary PRACE number