Minutes of Geant4 Steering Board Meeting of 2 July 2009
2 July 2009, 08:00-11:00 CET
Participants: V.Ivantchenko, J.Apostolakis, G.Folger, S.Banerjee, G.Cosmo,
P.Cirrone, S.Incerti, M.Asai, H.Kurasige, K.Murakami, J.Perl, D.Wright,
A.Ribon (GRID testing only).
- New SB members: Paul Gueye (JLAB), replacing G.Montarou.
- Minutes SB meeting April-2009 approved.
41.2 - Closed. See report by email by T.Sasaki.
42.3 - Open. No mail sent yet. Can submit tags ...
51.2 - Closed. See mail sent by D.Wright and web page.
51.3 - First draft available; follow up action to contact K.Amako.
51.4 - Open.
51.5 - Open.
52.1 - Open.
52.2 - Closed. No impact for minor release.
53.1 - Open. Some progress made, tests will be soon available.
53.2 - Open.
News/Reports from meeting
- IAEA Spallation meeting, Vienna 4-8 May, Dennis reported.
- IAEA CRP meeting, Catania 7-11 June, Pablo reported: meeting had the main aim study of data for hadron-therapy medical applications. 2nd meeting. One responsible per code packages participating (Geant4, Fluka, ...). See report linked to the agenda.
- J.Apostolakis comments that presentations to external meetings/conferences should be first discussed within the Collaboration (EM and hadronic groups), at least one month in advance.
- Testing: went fairly smooth; first release completely done using the new testing system on the nightlies; now fully operational. From the Bonsai DB, at the top, there's a link to the results of the nightlies; link to Tinderbox is now obsolete and will be removed. It would be good to have a link also from the STT pages. Requested short lecture on the nightlies at the Catania Workshop.
- Asked to A.Ribon to shortly join for discussing the GRID testing: very stable, no crashes; on the technical side, run on SLC4 (32/64 bits) and SLC5 (64 bits at CERN), run also on the OSG US GRID; no longer building on the site, but sending pre-built binaries, using environment configuration for choosing the physics configuration. Changed quite a lot the AIDA setup to have SLC5 working, future plan to move to use directly Root. Lists tested included the usual set; new lists have been tested only locally, QGSC_CHIPS rather stable, not the case for QGSC_QGSC. No changes in CPU performance with 9.2.p01; reproducibility has improved, violations are decreasing by a factor 2 respect to 9.2.p01 (violations sometimes appear when comparing jobs run in two stages vs. single run). Valuable help received from the GRID deployment team.
- Changes: multiple-scattering revisions, CHIPS proton extensions, ...
- EM Validation carried out for both production and EMV physics-list is identical to 9.2.p01; MSC in production has not been changed, only the development version was revisited for the description of tail distribution, to make it competitive with other models, like Penelope MSC.
- Physics-lists: new lists, two with _CHIPS using the new functionality in CHIPS, they're not part of STT testing yet, Hadr00 test will be used for this and also the calorimeter validation test. P.Cirrone asks if possible to rename physics-lists to more reasonable tag-name for medical physics... something like "hadron_therapy" or similar.
- GDML also extended to support user extensions for the writer.
- Preparation for 9.3: improvements to the GRID testing procedure to be implemented by then with better automatism and standard setup for analysis.
Geant4 Workshop 2009
- Facilities: same location of last Catania Workshop, but new site in the laboratory; also facilities to have lunches/dinners; big main room for plenary sessions, as well as lunches; smaller rooms in the laboratory for parallel sessions. To be decided if to have extended plenaries until the coffee-break in the afternoon, or follow the traditional scheme of plenaries only in the mornings.
Unless there is a strong need, we should give more time to the parallel sessions, i.e. keep the traditional scheme of afternoons dedicated to parallels, possibly three full afternoons.
- Social dinner on Saturday.
- Collaboration sessions starting from Monday.
- Parallel sessions requested: Google sheet distributed. No contribution received yet. SB members invited to fill it in !
Discussion in the working groups is requested in the next weeks and a reminder email to the SB should be sent.
- Asked for volunteers for the two parts (Collaboration and Users parts) , J.Perl will contribute.
- Internal Workshop: themes are, speed, hadronic modeling and validation, EM validation, architecture.
- A session on multithreading should be considered, maybe a plenary ?
Issues from Users
- Space Users Workshop, Madrid in May:
- When creating nuclei with excitation level, Vanderbilt users were unable to simulate for exceeded memory capacity or time.
- Successor for XMM and Chandra ongoing, created model for X-ray reflection for use in creating such design. Possible contribution classes to Geant4 code base. May be contributed in EM modules.
- Fragment production in hadronic interaction: when count of fragments is produced, Binary cascade and other models agrees, but for energy deposition model responses are quite different in spectra.
- Non-ionizing energy loss should be simulated separately.
- Technical Forum: expected one in the next weeks or in September.
- ATLAS hadronic calibration workshop: J.Apostolakis will discuss details with the hadronic working-group.
Delta Review 2009
- Proposed to agree to make report public, and request this to the OB.
- Live link to the Google document should be provided. M.Asai provided it to the SB members.
- Asked to assignees to complete the form regarding planning.
- Proposed task force (by J.Apostolakis) : J.Apostolakis, M.Asai, G.Cosmo, T.Sasaki assigned for assessment.
- X.Dong has been at CERN for the last 3 weeks; 24-cores system shows good scaling up to 8-10 threads, poor scaling beyond that. 15-24 threads show poor performance and high contention (locks?); Xin has updated the prototype to 9.2.p01. The task-force will have to assess if/how to provide changes in the code in the code base (some changes are not trivial).
- It should be also useful to know who will be ready to use multithreaded Geant4 and make their application thread safe as well, or if other venues will be followed.
- New fellow started at CERN as of July 1st, Andrea Dotti, working on hadronic physics validation. To be proposed as member of the Collaboration by D.Wright.
- Tutorials: CERN (beginners plus some intermediate level, reuse material from the previous LAPP course, M.Maire will contribute; updates to lectures are welcome if any), SLAC, Catania, and KEK; another is planned at Gran Sasso for INFN people, but to be confirmed after earthquake.
Requested pages for advertise the courses as soon as possible.
S.Incerti informs that updated slides for the low-energy part will be provided by September.
- Concern expressed for advertisement of courses not agreed by SB. Minimum requirement is the courses must be open, notified and accepted by SB. Similar rule must apply for presentations about Geant4 at conferences.
- External training courses or closed course can be posted to the training page, but marked as external; closed courses should appear only in the training page.
- How to react for events/papers/presentations organised without following the rules foreseen by the Geant4 Collaboration ? Affiliation to Geant4 should be asked to be withdrawn to the journal of the publication, for example.
- Regulations for addressing publication issues should be completed, possibly by the Geant4 workshop in Catania for discussion in the SB.
Asked to those who raised the issue to participate to the discussion and to propose revision of regulations.
Action: J.Apostolakis to propose revisions to the regulations to address publication policy and revision of 2006.
Meeting closed at 11:05 CET.