July 17th 2013
Attendees: John Allison (OB Liaison), Michel Maire, Sebastien Incerti, Peter Gumplinger, Makoto Asai, Marc Verderi, Luciano Pandola, Joseph Perl, Andrea Dotti, Koichi Murakami, Gabriele Cosmo, Gunter Folger, Alberto Ribon, Dennis Wright
Excuses: Ivana Hrivnacova, Daniel Elvira, Hisaya Kurashige, Vladimir Ivantchenko
- Regular items
· Open requirements, pending actions, bug reports and internal development issues are reviewed.
· Problem reports: If the user reporting the problem is not responding to the request of additional information for more than 3 months AND the reported problem cannot be reproduced, that problem could be classified to INVALID. Regardless of users further response, if the problem is reproduced, the report is kept and addressed.
- Towards 10.0 release
10.0-beta release was delivered on time.
Already good performance has been measured and reported to users.
- Summary of 10.0-beta release and its outlook
· Useful documents for developers and user’s code migration are accessible here:
· A preliminary benchmark with full CMS geometry on 8-core Intel Xeon hosting one Intel Xeon Phi (60 cores) card showed a very good linearity up to ~200 threads (until the point the card’s memory was exhausted). Memory consumption per thread has been measured as ~40 Mb/thread, starting from a base of 200 MB data shared. Aiming to further reduce it considerably by the public release.
· Physics performance is statistically equivalent with sequential mode. In 10.0-beta, event reproducibility is guaranteed at 91% level (99.9% after a fix introduced on top of 10.0-beta). The remaining cases are likely due to known issues in the treatment of ions, which will be tackled in the next weeks.
- Source code
· A reminder will be sent to developers to clean-up code after automatic conversion applied for migration to multi-threading, with a short recipe for addressing the common cases.
· Major use cases should be illustrated by at least one example properly migrated to multi-threading.
· TWiki pages for multi-threading will be integrated in the users guide.
· The simplified-calorimeter test has been ported to multi-threading. Need a second validation application preferably covering different geometry/physics (e.g. low-energy EM physics).
· It was noted the need for tools to detect MT-unsafe features/practices: this is now becoming high priority.
· It was agreed that one more beta is not needed, given also the time restrictions; development/candidate releases will be provided to major customers as needed.
- 2013 collaboration meeting
· Indico page for preliminary agenda:
· Session titles and chairs will be fixed in the following weeks.
New publication board member
· Daniel Elvira will replace Dennis Wright in October.
· Pablo Cirrone is the new publication board chair.
- Full paper for MC2013
· A general invited paper for MC2013 is under preparation.
- Demo and posters for MC2013
· Up to two desk/laptops for demo can be set up. One could be slide show, other running applications.
- New general paper
· A new general paper for a journal publication is under preparation. It is targeted for early next year. A solidified outline has been already circulated. Will call for section writers, likely working-group coordinators
· Granular papers are encouraged. They will be referred in the general paper.
- Recent events
· PTCOG, Essen, Germany, June 2-8th
Geant4 well represented in the domain.
· Geant4 Course at the 10th Seminar for Nuclear, Sub-nuclear and Applied Physics, June 2-8th
Software seminar running every year and hosting a Geant4 course since few years. There were about 20 attendees at the course (more at the seminar).
- Near future conferences
· 2013 Geant4 user conference at the physics-medicine-biology frontier + tutorial, Bordeaux - October 7-11th
60 abstracts received, already full!
· MC2013, Paris, October 27-31st
MT abstract has been accepted for oral presentation
Need to check the status of INCL and EM abstracts
· IEEE/NSS/MIC/RTSD-2013, Seoul - October 27th - November 2nd
One oral contribution accepted for hadronic physics
· CHEP, Amsterdam, October 14-18th
One oral contribution on 10.0 release and paper + 3 posters accepted
· SuperComputing 2013, Denver, November 17-22nd
Wednesday, July 17, 2013 - 12:00