December 7th, 2012
Attendees: Mike Kelsey, Dennis Wright, Vladimir Ivantchenko, Ivana Hrvnacova, Luciano Pandola, Peter Gumplinger, Koichi Murakami, Sebastien Incerti, Gunter Folger, Marc Verderi, John Allison (OB liaison), Joseph Perl, Andrea Dotti, Pablo Cirrone, Daniel Elvira, Alberto Ribon
Excuses:Takashi Sasaki, Hisaya Kurashige
- Reports from tasks
1.1 Computing performance
· 1 - Profiling/benchmarking of Geant4 9.6 release
o Two candidate versions previous to the release were profiled.
o The overall performance of 9.6 is comparable to that of the previous reference tag 9.5.ref09.
· 2 - CMS issue with the new glibc math library distributed with RedHat
o Investigations made in CMS for the High Level trigger software have reported a slow down in CPU by at least 5%; similar results for their reconstruction software.
o The reason for this performance degradation is in a ‘fix’ made for floating point rounding and exception handling (feraiseexcept in mathlib). The G4CP team estimated the impact of the new glibc library in RedHat running its performance/benchmarking tools on the SimplifiedCalorimeter. It was measured a performance degradation of the order of 15-30%.
· 3 - Plans related to release 10 (Geant4-MT) profiling
o The G4CP team has developed an application, cmsExpMT (multi-threaded version of cmsExp - a standalone version of the CMS full simulation software). cmsExpMT will be used for profiling/benchmarking 10 reference releases.
1.2 Physics validation
· A prototype tool to generate on-the-fly plots from the FNAL DB has been created.
· It was verified that Geant4-MT gives same results as sequential version (reproducibility: starting from same random seed, obtain same results).
- Regular items
- Open requirements, pending actions and open bug reports have been reviewed.
- Reports from recent events
3.1 Technical Forum
· Captured major requirements will be recorded in the requirement tracking system.
· Request from Mu2e: to have online what bugs affect what releases. Would require queries in Bugzilla to be filled properly. Would require fixed bugs to be documented in Bugzilla, even if fix is made quickly.
- 9.6 Release
· Note: the development release and the released public code are not exactly the same: shadowing of variables checks by the compiler is not activated in the public release.
· Inclusion of G4SystemOfUnits.hh and G4PhysicalConstants.hh should not be done in .hh (and .icc) files but in .cc only, where necessary. In header files and inline code one should explicitly include the CLHEP headers (CLHEP/Units/SystemOfUnits.h and CLHEP/Units.PhysicalConstans.h) and use the CLHEP:: namespace for each symbol used.
- Steps towards Geant4 version 10
· The conversion to MT for release 9.6 will be the last one performed. Any subsequent development to the SVN main trunk will have to guarantee thread-safety.
- Documentation and web pages
· Agreed to have the documentation WG involved in the review and maintenance of the common web pages.
- 2013 Collaboration processes
· Registration to SVN will start at the beginning of January.
· WG coordinators are requested to complete the WG census by the end of January.
8.1 Near future events
· Geant4 Space Users Workshop
· 2nd Geant4 Australian School and Monte Carlo Workshop
8.2 2013 Collaboration meeting @ Seville, Spain
· No particular conflict was identified for two proposed weeks of Sep. 16-20 and Sep. 23-27.
· CHEF2013, Paris, April 22-25
o Abstract submission to start by January 1st.
o Presentation and presence to this conference encouraged.
· 2013 Geant4 user conference at the physics-medicine-biology frontier + tutorial, Bordeaux - October 7-11
o http://geant4.in2p3.fr/2013/ (preliminary)
· MC2013, Paris, October 27-31
o Deadline is 15th of December.
· IEEE/NSS/MIC/RTSD-2013, Seoul - October 27 - November 2
Friday, December 7, 2012 - 12:00