Minutes of SB meeting December 2009
Date: Friday, December 11th, 2009
Time: 07:30 CET, 15:30 JST. Duration: 2 hours
Editor: J. Apostolakis
Notes: First draft Dec 11th, 17:30 JST
Revisions: Wed Dec 16th, 13:15 CET
- Regular items - 10 mins
Approval of minutes of SB meeting of October 17th - 22nd, 2009
Minutes approved with one change:
- Notes on at risk items to put into separate document and summarised.
Pending Actions of October 2009 reviewed and updated
Release 9.3 - 15 min
- cand-05 Fri 4 Dec given for testing (integration, regression). Also distributed to experiment representatives for initial testing and feedback. First feedback from CMS is positive - most problems have been solved.
- Additional tags with fixes identified by regression testing. Will be included into the first version of the reference-tag. At least the correction for the standard license text, std:: methods will be added. Then packaging for the release will be undertaken.
- New version of CLHEP (220.127.116.11) with fix for RANECU on 64-bit.
Nearly all contributions have been submitted
- One is promised for today (the deadline)
Preparing final version during week of 14th Dec
Many incremental changes - no big changes
- testing of tags on top of last candidate
Identified additional issues in Binary Cascade
One or more platforms have no platforms failing
- other platforms with only compilation warnings from non-G4 code
Report from Grid Regression Testing (A. Ribon / A. Dotti)
- Report made - Separate document
Open issues - actions underway or needed
Preparation of Geant4 2010 Architecture Review - 20 mins
Refinement of proposal for Review Mandate & Scope
Proposed Schedule - 2nd draft
- CPU performance and memory use,
Proposed migration to SVN service - 10 min
Timescale of cutoff of CVS service at CERN
Many projects have not planned to migrate
Geant4 is developing and assessing plans to migrate
Not committed to move from CVS to SVN
Is it suitable for our development model ?
Ability to have a sticky tag is essential for G4
- Action (Gabr & Gunter) Report if this is solved
Expect CVS service to continue until December 2010 at least
Need for fallback plan or alternative
Migration of CVS service could be undertaken
- GF: Part of tag collection would need to be migrated as well
Plan deadlines for final report on issues and decision
- Plan for decision in June 2010 regarding code repository
Potential trial SVN service
Only once a mature proposal is prepared
- Purpose ? familiarise developers
Collaboration Workshop (Users and Collaboration Meeting)
Potential Dates: 4th-8th October ?
- Conflicting meetings: Monte Carlo 2010 (MC2010) - Oct 17-22
Old proposal: ESTEC (4 day Collabor.) & CERN (3 day Users Mtg)
Revised proposal ?
- 5 Day Collaboration Workshop @ ESTEC
Action (John) Report by next SB on confirmation of location
- Prepare plans for future Users Workshops (ready for announcement)
- Working Groups are free to organise side meetings
Contributions to conferences in 2010 - (Note the deadlines!)
Physics for Health in Europe @ CERN
Meeting: 2-4 Feb 2010
Abstract deadline: 8 Jan 2010
- SB endorses that Seb Incerti or Petteri represents G4 @ this meeting
MC2010 (Tokyo Oct.17-20) - 30 January 2010 (submission)
- Responsible - Makoto
CHEP 2010 (Taipei, Taiwan, Oct. 18-22) Deadline ??
Responsible ? Anyone attending ?
- Identify Responsible from Japan - Koichi ?
IEEE NSS 2010 (Knoxville, TN, USA. Oct.30-Nov.06) Deadline ??
- Responsible: Dennis (or whomever he convinces)
Coordination of submissions: Proposal for process (below)
Preparation of 2010 Collaboration processes
- Geant4 census(es) (target date: 14 Feb)
- Election of SB representatives (target: March 30th)
- Election of Spokesperson (start: April 15th)
Progress with Review 2009 response(s) - open issues
- Unresolved 2009/priority issues
- Planning future (2010+) action
New proposed Collaboration members
Dennis: Mike Kelsey (hadronics) - approved
Sebastien: propose new student to work on G4 DNA needs access to CVS
- Reference tags only required.
Pablo sent list of potential contributors in advanced examples
Remind that members have to submit their G4-related to the SB for approval before submitting it for publication.
Question/Concern from Takashi
- What kind of examples should be distributed by Geant4?
Ensure that WG members are involved in development or testing
- Case: contributing model
In other WGs (Low E, Std E EM)
Space Users Workshop 2010, Aug.18-20, Seattle, WA, USA
- Decision to rollback code to June 9.3 beta release was agreed.
- Agreed on need for a process to address remaining matters and to establish practices for future.
Reviewing non-advanced examples (postponed for next meeting)
===== End of Meeting 17:35 JST ===================
1. Proposal for Process on Conference Contributions in 2010 (and beyond)
Itemise key topics/potential submissions for each conference:
- ensuring best targeting and good representation
- use collaborative tool (eg Google Document) to share proposed submissions
- All contribution (from general, to WG-only, to few authors)
Circulating proposed contributions at least 2 weeks before deadline
(Addition to proposal) Responsible to collect abstracts for each meeting. ( Agreed and Makoto agreed as person for MC2010)
WG coordinators collect proposals in their working group
Need also to encourage submitting on particular topics
- Identify key topics for conference
MC-2010 deadline is 15 Jan, Makoto expects constant communication and early draft before this.
Question: author list ( clarify process )
SB is responsible & can delegate
- WG is responsible for smaller ..
2. Proposal for Geant4 2010 Architecture Review
The proposed action for 2010 is an internal review of the architecture of Geant4. Redesign work will be undertaken as a result of the review.
Planned outcome(s) of Review:
- a prioritized list of areas for redesign, with effort estimates
- New assessment of technologies for impact - recommendations which to prototype
This will be a review of priority areas:
key interfaces, in particular those between categories (Process, Track, Step/Pt, ParticleChange, Touchable, PhysicalVolume, PhysicsTable, Material, Particle)
- Process sub-issue: process proposed status (forced: 3+ variants)
- Should PhysicalVolume ( & all geometry) become read-only (currently R/W) ?
external interfaces, including all dependencies
- 3-vector, Lorentz vector, rotation-matrix, ..
- including (G4)Exception, RNG-engine, PhysicsTable, PhysicsVector, ..
assessing the use relationships between categories (reduce crosstalk)
impact of addition of ‘major’ functionality since 1997
- including cuts per region, scintillation, parallel navigation, ..
- potential impact of future new functionality (e.g. multiple density, propagating phase for strong/weak interactions ? ) - find which affect architecture
Examine major choice of ‘dumb’ bucket for track, step, volumes, …
- Clearly this is a fundamental issue of the architecture of Geant4.
Consider technology shifts such as multithreading, multi-core, new computer platforms / architectures, (memory hierarchies, mixed CPU-GPU, other?), grid.
Target areas for improvement
Potential targets for future improvements
CPU performance and memory use,
- Ensuring that design promotes/ensures correct use of interfaces
adaptability to future platforms (multiprocess, multithreading, .. )
uniformity of the use of the interfaces classes and other infrastructure classes ( issue of design or implementation ? identify the key issues )
Weigh benefits, risks, effort required for potential improvements.
Review to report in time to use the 2010 Collaboration Meeting (Sept/Oct 2010) to plan the architecture revisions.
Propose potential date(s) for a mini-workshop (action: organise date via doodle ? )
> Target June 2010 (fallback May 2010)
> Some focused sessions (Architecture
> Most would involve SB member (as many as possible - try to ensure that key areas are well represented at meeting. )
Preparation Meetings for mini-workshops
Jan/Feb 2010 - May 2010
First face-to-face meeting: 22-26 February 2010
Second face-to-face mtg: late March / early April (Japan?)
Depending on the topic all relevant Category Coordinators will be invited and expected to attend.
Technologies and Methodology
New/Issues (JohnA, 25 Nov 2009, revised in Tsukuba, 7 Dec 2009):
New C++ language features (exceptions; STL; .. ; C++ 0x ? )
impact of developer understanding of features needed for many developers
using advanced features sparingly (so that a fraction of developers is exposed.)
- Template meta programming is not well understood
adopting policy to allow, forbid or encourage advanced features.
New programming technology
- why do we have different implementations of singletons? Is it not a burden?
- are there any other ‘low-hanging’ fruit for future improvement ?
How do we focus on key elements and reach conclusion about C++ features, tech?
- Focus on Criteria for use of new language features or technology
‘Width’ of use of new language features/technology. Choice between
- use by few (core) developers only (how many?)
- use by many developers
- exposed to users.
Assessing constraints today and forecasting it in 3 or 5+ years time
- Evolution of technical constraints (memory, CPU architectures, .. )
- Programming knowledge constraints (coping with few experts amongst many less capable programers)
Focus: identify issues
Friday, December 11, 2009 - 12:00