Public minutes of 76 th Steering Board Meeting

February 6th, 2013

Attendees: Makoto Asai, Takashi Sasaki, Sebastien Incerti, Marc Verderi, Andrea Dotti, Koichi Murakami, Ben Morgan, Ivana Hrivnacova, Dennis Wright, Joseph Perl, John Allison (OB liaison), Gabriele Cosmo, Gunter Folger, Michel Maire, Alberto Ribon, Vladimir Ivantchenko, Luciano Pandola

Excuses: Mike Kelsey, Daniel Elvira, Hisaya Kurashige

Regular items

1.1 Open requirements

o http://jira-geant4.kek.jp/secure/Dashboard.jspa

o Open requirements are reviewed.

o New requirements captured from the Technical Forum in December 2012 are discussed. The following requirements have been accepted and listed to the requirement tracking system for further examination:

o Allow for use of alternate (fast) mathematical libraries;

o Allow for suppression of the NeutronHP package warning messages on user’s request;

o Indicate when Geant4 does (not) take ownership of pointers;

o Physics list factory extensible with user’s physics list or to allow for combinatory with user’s specialized physics module;

o Change of at rest logic of the stepping manager to allow stopped tracks to be accelerated and further tracked;

o One requirement, on hadronic interaction of short-lived particles (B and D meson) needs further information.

1.2 Pending actions

o The Steering Board action items are reviewed.

1.3 Open bug reports

o http://bugzilla-geant4.kek.jp/buglist.cgi?query_format=advanced&bug_status=NEW&bug_status=ASSIGNED

o The open bug reports are reviewed.

o Note:

o Once the new JIRA system for developers will be made available, the internal bug reports on under-development code will be transferred to this new system. Collaborators are requested to use this new JIRA system for their future internal reports of development issues.

o As announced earlier, the assignee of each bug report is requested to assign to the report the Geant4 version corresponding to the actual introduction of this bug.

Patch release

o The patch release has been announced today, details are provided in http://cern.ch/geant4/support/Patch4.9.6-1.txt

o This patch includes in particular the important fix on the interaction between the multiple scattering and the Transportation.

o It does not include yet a fix to problem report #1397 affecting the AtRest processes logic in the stepping manager.

2013 collaboration census

o The census is almost done, few people are still being contacted.

o The Collaboration member list and mailing lists will be updated accordingly.

Progress in Multi-Threading (MT)

o A report on the meeting at Boston dedicated to this conversion is made, and the status of converting Geant4 to multi-threaded is discussed.

o The following points need to be addressed:

o The multiple-scattering process has been changed in the 9.6 patch. It works with the MT version, but its lack of table sharing makes it not optimal w.r.t. to memory usage.

o The static and global variables are converted to Thread Local Storage (TLS) variables by default, which is not optimal as TLS variables induce some overhead. A code inspection and cleanup is necessary to adopt such TLS variables only in needed cases to help in recovering part of the performance loss in sequential mode.

o The CLHEP library was transformed as well in G4-MT-9.5.p01 by mistake which should not be for release 10.0. For 10.0 the reimplementation of in particular the random number generators, in the way it was done in G4-MT-9.4.p01, will be done.

o Some issues will require further efforts:

o Windows support: static variables in inline template function cause problem;

o Split classes: some classes need to be split into read-only and read-write classes. Here also the automatic conversion from static to TLS done by default is not optimal and requires review.

o Working Group specific items for MT conversion:

o hadronics : to remove unnecessary TLS

o EM electromagnetic : multiple-scattering process

o geometry: navigation and field to be reviewed

o run : new run manager and related classes

o visualization : one visualization scene from all threads

o all : review use of TLS

o Some examples will be converted to multi-threaded by the end of February.

o A penalty of the sequential mode is observed with the prototype and needs to be understood

o currently, ~15% of the time is spent in redirection with TLS.

o Windows version is not functional at the moment.

o inline code for static data are not allowed by Windows

o planning to restore the sequential build on Windows by the next development release

o Instructions for maintaining thread-safety will be distributed.

Improvements of documentation

o The Toolkit Developers Guide is not maintained since many years.

o The Steering Board agreed to revitalize the effort of maintaining/reviewing it.

o The conversion to MT will need documentation on the key classes with shared data.

o The manual is not aimed at providing detailed things but just global view. The purpose of the manual is to provide expert users with relevant information on how to extend the functionalities of the toolkit, providing clear guidance (e.g. How to add a new physics process (EM or hadronic), or cross-section, integration in existing physics-list or addition of a new tailored one; addition of a new solid; specialization of the navigator; etc.).

o Documentation WG will conduct the collaboration-wide efforts.

7.1 2013 Collaboration meeting

o Date is fixed – September 23-26.

7.3 Near future conferences

o CHEF2013, Paris, April 22-25

o http://llr.in2p3.fr/chef2013/

o deadline end of February for abstract

o ACAT, Beijing, May 16-21

o http://acat2013.ihep.ac.cn/

o 2013 Geant4 user conference at the physics-medicine-biology frontier + tutorial, Bordeaux - October 7-11

o http://geant4.in2p3.fr/2013/ (preliminary)

o registration to be open by end of February -> early July