Minutes of 10 Mar 2009 meeting of Geant4 Technical Forum
Editor: G.Cosmo, 10 Mar 2009
- Starting at 16:35.
Update on Open Issues
+ #1503, low-energy physics-lists: plans in 2009
+ #1202, RPG model under development, missing resources
+ #1102, included in release 9.2; proposed to close
+ #704, data provided for all available; proposed to close
+ #0304, 0502, 0601, 0605; closed
+ #0303, ongoing process improvement; closed
+ #0106, regression tests now part of standard release process; propose to close
+ #1501, fix released in Feb-2008. Proposed to close
+ #1302, are handles to PL and processes still required ?
+ #0402, ongoing progress; still open, need to extend coverage of processes
+ #0309, no progress
+ Request for underground experiments: no progress yet, need to prepare an on-topic
discussion for the next meeting
Geant4 Relase 9.2 - and upcoming patch
+ Question on performance tests applied on the new models
+ Easy transition to 9.2 reported by T.Roberts; appreciated the introduction of
the physics-lists factory.
+ Still requesting better documentation of the physics-lists
Proposed Geant4 development plans for 2009
+ No feedback received before the meeting, apart from a request which will be
presented lated in this meeting.
+ Rewriting of the parameterised model (RPG) is postponed likely to 2010.
+ Muon MSC, model is based on Wentzel approach; want to improve accuracy and possibility
to combine with the single-scattering. Agreement with MUSCAT data is now much better
in release 9.2 of what it used to be; plan to provide a variant of the physics-list which
will be optimal for this use-case.
+ Requested better description (just a short description) of the various physics
configurations; already available as one of the requirements discussed above.
+ MSC development has been ongoing for long time, no revolutions is expected from what
planned. MSC processes are now divided per particle-type, for muons and hadrons the
traditional default models are kept unchanged in good part.
+ Reaction from the users community is expected in the next couple of days.
New requests, requirements, priorities, issues
+ CMS: Next major release 3_1 expected for March 28th, should be the release for the "mass"
production; based on 9.2 (+patches). After 3_1 should exploit the data-driven improvements;
required to base to 9.2.pXX until spring/summer 2010 with bug fixes not impacting physics,
but eventually performance improvements. Willing to further improve Geant4 in the longer
Requirements: use of physics-lists per geometrical region, would benefit mixing of different
calorimeters (ECAL/HCAL), allow optimal choice of physics (QGSP_BERT_EM vs QGSP_BERT).
Improvements in the tracking in field, local field manager per region and solve the open
issues with the visualization of Boolean shapes.
-> First comment by J.Apostolakis: development related to physics-lists per region can
certainly not happen in the 9.2 release series.
+ ATLAS: No crashes in production since October 2008; 250 millions events, with both 8.3
and 9.1.p03. Only a minor private bug fix for the counting of final state particles.
Working on the migration to 9.2, plan to use in production in fall; change in sampling
fraction in tile-calorimeter by several percents; fall-back to 9.1.p03 is a possibility.
Using 32-bit builds in gcc-3.4.6; planning to go to SLC5/gcc-4.3 still 32-bits (not
supported for 9.1.p03).
-> Also CMS confirms the stability in running Geant4.
-> LHCb notice an increase in hanging events using 9.1.p01 (15-20%).
+ TPC simulation: T2K experiment. In 9.1/9.2 resolution of 10%, changed from the 6% which
used to be in previous versions (8.2). Reproduced with example TestEm5.
Discussions in the Geant4 Forum and private discussions; problem seems related to the
universal fluctuation model; an attempted fix provided by V.Ivantchenko for the fluctuation
model when step-limitation is applied seems to provide positive results for this case study.
-> Rather specific case which should not worry LHC physics. It is only related to TPC
simulation; could affect ALICE though.
Fix is under study and will not be included in 9.2.p01.
-> Can it have effect for step limitation in silicon ?
Can only have effect if gas is medium.
-> Universal fluctuation was modified considerably since release 9.0.
Old code could be used in 9.1/9.2 series but physics lists should be custom.
Support for Releases in 2009
+ Proposed to fully support 9.2 until November 2009, including fixes.
-> CMS and ATLAS require support until end of 2010 (end of data taking).
-> Should be clarified what type of problems should be considered in the support.
Will be discussed in the Steering Board and feedback requested.
-> Scenario also for level of support for 9.1.p03 in case of fall-back.
Supported Computing Platforms for 2009
+ SLC4 to be dropped by December 2009 and no longer supported from the future release.
Will also include drop of gcc-3.4.6.
+ SLC5 support including both native and gcc-4.3.2 compilers; icc-11 first-level support.
+ VC++ 9.0 on Windows
+ MacOSX 10.5, gcc-4.0
-> Future versions of MacOSX will be tested as first-level as they come out.
Report from Delta Review
+ Report available but not yet public. Feedback of closeout available with positive
feedback with few points for recommendations. Emphasis on the level of progress
made by the Collaboration in the different areas.
Closing meeting at 18:23.
Tuesday, March 10, 2009