Simulation Project Leaders Meeting 16-November-2004 ==== Attendees Paolo Bartalini (PB), John Harvey (JH), John Apostolakis (JA), Witek Pokorski (WP), Alberto Ribon (AR), Paola Sala (PS), Gabriele Cosmo (GC). --------------------------------------------------------------------- ==== Minutes LHCC review: not open to public but Simulation Leaders can partecipate. --------------------------------------------------------------------- ==== Generator services - Paolo The answer to the reviewers (Generator Services internal review of March 2004) has been prepared. It will be finalised by Monday, November 22nd. Roles for the librarian and people involved in the release integration are now better defined. Alberto Ribon (LCG) works on 25% as MC expert in WP1 and WP2 (OO packages, ThePEG respectively); he will share the "liason" activity with the MC authors and LHC experiments together with PB. Andreas Pfeiffer (CERN) will dedicate 25% of his time for acting as the GENSER librarian and technical support to releases. Having Andreas as librarian will allow russian contributors to work more on development and validation as integration testers. New manpower resources will be available from LCG Russia in 2005. The new balance corresponds to 1 FTE in GENSER and 1 FTE in MCDB, (0.25 FTE increase for both activities). There will be always one person at CERN for each project. Rotations will be performed among a restricted group of MC experts already trained on GENSER. There will be few intersections between people allocated for WP1 and WP3 corresponding to resources from different institutions. An internal meeting of LCG Generator is scheduled for November 17th and dedicated to discuss the release procedures in order to define the release policy document (that will be edited by Andreas). The next public LCG Generator meeting will be held on November 25th and will be dedicated to HIJING validation and EVTGEN integration. Progress in the developments of the LHC version of EVTGEN and attempts to unify it will be reported. - GENSER (WP1): Mikhail Kirsaov (Troitsk) is the current integrator/tester. He will finish his shift on November 25th. The work of Mikhail is resulting in the production of a test version of GENSER still following the older policy: GENSER_0_3_0. Access to GENSER_0_3_0 has been granted to the beta testers in the experiments. It supports SCL3 and, on top of new versions of the already inserted sub-packages, it contains MadGraph and MC@NLO. The CASCADE sub-packages is still not released due to its dependency on the CERNLIB. Sergei Slabospitsky (IFVE) will be the new integrator/tester and is currently having a one month "overlap" shift with Mikhail. He is a MC expert and author of TOPREX. He will take care of the insertion of TAUOLA, TOPREX and other small size MC packages in GENSER. Alexander Sherstnev (MSU) worked during the LHC/HERA workshop for the integration of SHERPA in GENSER. There are still problems to address concerning the definition of the SHERPA Workspaces in GENSER. Sharing of the binaries between different users seems to be an issue. This would give serious problems to use it in the standard production of the LHC experiments. A solution to fix such problems was proposed and is currently being evaluated. Sergey Makarichev (ITEP), also covering the 3 months shift for MCDB from November 22nd, will dedicate 30% of his time to SHERPA integration. Sergey already collaborated with the SHERPA authors in the occasion of his first visit at CERN (summer 2003). - VALIDATION (WP4): There's a general agreement in a wide community (including most of the LHC experiments and other independent projects like JETWEB) that LCG Generator should play some role in coordinating an effort for unifying the way tuning and physics validation is performed, avoiding possible duplication of the work. Dedicated discussions on how to organize this possible task force are currently ongoing. One of the main experts on the experimental side, Rick Field (University of Florida), is also involved. PB was invited to present the recent developments on generator tuning and physics validation to the Monte Carlo workshop that was held during the CMS/CPT week. Dimitri Bourilkov (University of Florida) produced an interface (LHAGLU) to improve the usability of the package LHAPDF from the most popular shower Monte Carlo, allowing to discard the obsolete PDFLIB. LHAGLU has been made available in LHAPDF from version 3.0 (included in GENSER_0_3_0). In general there was a very good sinergy between all the actors involved in the LHAPDF validation process. JINR has an independent activity on hadron-hadron, hadron-nucleus and nucleus-nucleus validation, including HIJING validation (contact person V.Uzhinski, web site http://hepweb.jinr.ru/). The organization of the project is very much along the lines of JETWEB. Periodic reports are given to the LCG Generator meeting. The next report is expected November 25th. In perspective, perservation of compatibility with the JETWEB tools will allow merging in the mid/long time scale. - MCDB & PRODUCTION FRAMEWORK (WP3): The main author of MCDB, Lev Duvko, is currently covering the MCDB shift. He mostly concentrates on the finalisation of the HEPML tags and on documentation. The new shifter (Sergey Makarichev) will concentrate more on the development of the search engine, authorization procedures (including GRID certificates) and support to additional browser (on top of the already supported one, i.e. Mozilla). At the CMS-generator workshop there was also a report from Filip Moortgat (CERN) on the development of the object riented LCG Generator level production framework (WP3). The resources currently allocated to the project, corresponding to 0.5 FTE, come from Oviedo and Santander. The framework is still at the design stage due to problems in defining properly the MC input (generalization of data cards, definition of a MC dictionary). Since ATLAS and LHCb have already a working solution, it is recommended to carefully study it in order to understand how it can be customized to the LCG needs. --------------------------------------------------------------------- ==== Fluka - Paola - Working on the public release of Fluka. Still manpower problems; Alfredo is currently very busy with LHC tasks and this may delay the final date for the public release originally foreseen for the end of this year. Work is going on especially for what concerns the elimination of the PEMF preprocessing and the neutron library where there are problems with the standard neutron data files. In parallel, work is going on the ion physics, restructuring of transport in the low-energy region and improvement on the user interface and I/O. - The Fluka manual will be published as CERN yellow report in December. --------------------------------------------------------------------- ==== Geant4 - John Deliverables: - release validation and acceptance suite: the goal is a comprehensive regression testing suite to monitor changes in key physics observables and guarantee (to the extent and for the cases possible) that physics performance is maintained. Eventually improvements can be documented, and potential degradations due to any changes can be corrected. The new suite simulates simplified setups, including approximate versions of the calorimeters in the LHC detectors. Currently work is focussed on tools for improving automation, so it can be run over many test cases, remotely and by different persons without requiring less expert knowledge until a final 'person' evaluation. The tool should allow to easily spot undesirable effects of changes, if any. Release 7.0 will be validated by using the first versin of this tool as part of the testing procedure. A meeting will be held this week to discuss in details the different techniques that can be used to implement this. Potential additional automated verifications are foreseen to cover multiple-scattering in thin foils, dE/dx in silicon layers, etc... - Currently in the release cicle for 7.0, scheduled for December 17th. The release will include new development, however for what concerns the physics, most of the developments were already included in the June release, and mainly refinement and improvements will be in 7.0. Some deliverables have been postponed to next year. Geant4 7.0 will support CLHEP 1.9 and will implement the migration to . Old compilers like gcc-2.95.2 and VC++6.0 will not be supported. - Usage of SEAL/MathLib: new dependencies to new packages would be an issue, also considering the limited usage that Geant4 makes of specific mathematical functions. --------------------------------------------------------------------- ==== Simulation framework - Witek - FLUGG: WP and PS have discussed about the possibility to extend Flugg for a more automatic handling of hits and detector sensitivity. It was common feeling that in order to implement something concrete and useful which could be also applied to test-beam simulations in a generic way, it is required additional manpower, evaluated on the order of 1 FTE for one year, where the person should have some experience in Geant4 and have knowledge of the Fortran language. With the current manpower available, extensions to Flugg can be provided with a very limited application (i.e. to simple examples, like those provided in Geant4 for the 'novice' set), provided that the interface will technically work. PS is also afraid that this could be misleading to the final users of Flugg. The development can in any case go in parallel with the development of the test-beam example or benchmark. JA makes notice that what has been done in VMC may help in this and should perhaps be considered. It was agreed that WP and AR will look at what is currently available in VMC and see if it can be applied for a test-beam exercise. - GDML: The Geant4 reader and writer are now ready and working. The whole LHCb geometry can be imported and exported with no problems; Writing out the whole geometry to GDML takes roughly 10 seconds, reading it in takes even less. A first exercise on importing/exporting the whole ATLAS geometry was done; the test went well in general but some problems raised, since in ATLAS there're cases of user-defined solids and some exotic usage of G4DisplacedSolid. The problem for G4DisplacedSolid can be easily solved and a fix for it is already under testing, while for the custom solids it will be necessary to provide in GDML a way for the user to customise the schema. WP also discussed with Rene about interfacing GDML to ROOT. Rene showed enthusiast to follow this work. The first step is to use the work made by Ivana Hrivnakova for interfacing Geant4 to ROOT, a first exercise made with it was successful (going from GDML to Geant4 and then to ROOT). There's a lot of interest from users (mainly from USA) about GDML; proposals for extending the schema have been received, in some cases also to consider persistency in a wider context, not strictly geometry-related (like sensitive detectors ...). These can be seen as possible extensions to GDML, however, GDML itself (core) will be kept just related to geometry descriptions. Radovan Chytracek is currently working to repackage GDML for the new version expected soon. - WP also exercised a reingeneering of the GDML reader/writer in Python. The new module became shorter in code of a factor 5. This can be seen as a step towards a general pythonisation of the simulation environment. Using this approach it may be possible to run a specific test-beam simulation taking existing geometry and sensitive detector setups with the minimal effort in interfacing. JH: will there be customers for a Python approach ? WP: it will give possibility to outcomers to test setups that would be otherwise difficult to prepare. Mainly for debugging and validation purposes. --------------------------------------------------------------------- ==== Simulation physics validation - Alberto - No much news since the last meeting. Experiments are currently busy with the test-beam data. - There will be a meeting on December 1st, with presentations by ATLAS on the linearity of beam precision and on EM calorimeter test beam in LHCb. - The possibility if to implement or not a third simple benchmark is under discussion: experiments are happy to have as many benchmarks as possible but are busy and cannot provide resources ... There's no much interest in continuing this activity from the Geant4 side, H.P.Wellisch would prefer to concentrate on more urgent issues related to test-beam data in the experiments instead. It will be asked to participants to the meeting on December 1st to come with ideas for possible simple benchmarks to implement and see if these can be of interest, in such case the work will have to be planned considering the existing resources. Manpower for continuing the benchmarking activity is anyhow seen as a real problem. JH: benchmarks should be more useful to MC authors to improve the simulation engine. PS: checking a program on complex benchmarks does not make much sense and cannot be considered a first check of the program itself. JA: it's an activity that should not be forgotten. - The publication for the pion-absorpion benchmark is currently pending. H.P.Wellisch has given feedback to WP; his point of view is different from the what was written in the current draft. The main issue seems to be related in agreeing about the possible effects that pi-absorption may have to e/pi ratio. Perhaps a solution is to leave to the reader to judge the results. - AR and WP will write a note about what has been done for the pixel benchmark, documenting the steps taken to interface with Flugg. The note could then serve also to start also with the calorimeter test-beam example. --------------------------------------------------------------------- ==== AOB Next meeting to be scheduled by mid-December.