Simulation Project Leaders Meeting 15-November-2005 ==== Attendees John Apostolakis (JA), Witek Pokorski (WP), Alberto Ribon (AR), Rob Veenhof (RV), John Harvey (JH), Gabriele Cosmo (GC). Excused: Paolo Bartalini (PB), Alfredo Ferrari (AF) --------------------------------------------------------------------- ==== Minutes - At the LHCC Computing comprehensive review held on November 14th, PB has presented the current status of LCG Generator Services subproject, with also focus on the recommendations advanced at the previous review in 2004. --------------------------------------------------------------------- ==== Simulation physics validation - Alberto - Meetings At the last simulation validation meeting (October 26th) there have been several presentations: from CMS on the HCAL test-beam 2004, where the known problems on shower shapes for QGSP on higher energies have been confirmed; the analysis was done using Geant4 6.2.p02 and needs now to be reiterated with version 7.1.p01 to verify possible improvements. For ATLAS, there was a presentation for HEC test-beam 2000-2001, where the response with three different versions of Geant4 was verified (5.2.p02, 6.2.p02 and 7.0.p01); no change was observed for the LHEP physics list, while improvements for QGSP have been confirmed in 7.0.p01. A presentation is planned at the November 30th meeting on the ATLAS combined EM/HD end-caps. There's good interaction between Geant4 developers and the experiments in this activity of verification of the shower shapes. The changes in the xsections introduced in Geant4 7.0 seems to have slightly improved the results. - TileCal 2002: no much progress in the generation of the new data sets; the original authors have been very busy in the last months. The tools are anyhow ready; progress has been made in the required Fluka extensions, where technical issues are being faced in order to identify a solution that can be valid in a general way, application independent. - There have been contacts with Tancredi Carli in ATLAS to try to use in a standalone mode his test-beam setup and preserve the data in order to use them for the long-term in the context of validation and regression testing. The idea is to eventually employ also other calorimeters test-beam data, and therefore allowing to have periodical cross-check with real data. - 3rd simple benchmark: some progress has been achieved; differences in xsections between Geant4 and Fluka which were originally noticed, are now fixed, they were due to a bug in the analysis; secondary interactions in Geant4 were missing in the analysis and were added. These fixed most of the discrepancies observed initially. Still to understand fully the selection cuts in the real data in order to reflect them in the analysis. Once this will be fixed, AR will contact the MC experts and the preliminary results will be presented. --------------------------------------------------------------------- ==== Generator services - Paolo (notes sent offline) - Events: + MC Truth meeting of September 28th (organized by WP) where usage of HepMC in the LHC experiments was also discussed. + LCG Generator meeting of September 29th dedicated to the release of GENSER_1_2_0 and MCDB demonstration. + LCG Generator meeting of October 20th; mini workshop dedicated to Monte Carlo validation (including Monte Carlo validation in the test package of GENSER). Large participation from experimentalists and theorists (in total around 35 participants) + Review presentation of LCG Generator at the LHCC computing comprehensive review on November 14th. + LCG Generator meeting of December 1st will be dedicated to HepMC. Author (M.Dobbs) and current CLHEP maintainer (L.Garren) have been invited. + Abstract for CHEP06 has been prepared & submitted. It concentrates on Monte Carlo validation (WP4). Speaker will be M.Kirsanov. - Manpower: + Currently no shifters. M.Kirsanov (GENSER) will join on Nov 15th; L.Dudko (MCDB) will join on Nov 15th. Previous shifters were Alexander Toropin (GENSER), Sergey Belov (MCDB/HEPML) and Anton Gusev (MCDB/HEPML). + INFN fellow (Marianne Bargiotti) is working at 30% on WP4 (MC Validation). + Research associate (T.Sjostrand) is currently away from CERN. + Only A.Ribon (liaisons, ThePEG), A.Pfeiffer (Librarian), and P.Bartalini (Coordinator) are currently at CERN. Since Andreas is moving to other projects, we should re-iterate the following question: who will act as new librarian for LCG Generator in the future ? - Milestones: + No new milestones to discuss since the last Simulation Leaders meeting and Q3 report, however, there's small update on some of them. + The procedure for packaging of individual generators is in place, with the release of GENSER_1_2_0 (October 20th) see: http://lcgapp.cern.ch/project/simu/generator/DownloadIndividualGenerators.html A test with a small bug fix release incrementing the third digit will be performed by the end of the year. - GENSER: + Release 1.2.0 has been made available October 21st; details are available in a note below. The new librarian tools demonstrated to be effective for early tracking of bugs in the pre-releases (Python scripts for package installation/compilation, post-installation checks). + Progress concerning the transition to g95 and gfortran (activity initiated in July by M.Kirsanov). g77, g95 and gfortran basically produce same results without optimization options (report from R.Yaari confirms our observations). Collaboration with Herwig allowed to solve problems with loading of Block data. Herwig 6.510, released on Oct 31st (i.e. too late for GENSER_1_2_0) is now completly cleaned up from this point of view. There's a strong request of ATLAS to have Herwig 6.510 in GENSER early December 2005 (it will be included in the bug fix version). - New OO MC generators: + Pythia 8.040 was released in July 2005. It is currently frozen until January 2006. Sjostrand claims that there are very few (if any) feedback from beta testers. It is stressed that this would be the best time to consider feedback from experiments. Beta testers in experiments are invited to attend LCG Generator meetings. They are encouraged to profit from the permanent forum provided by these meetings, also attended by the Pythia 8 author. + Herwig++ should produce the first usable hadron-hadron release at the end of the year 2005 (very limited functionalities will be supported). Significant progress are reported, at the same time authors think that it will be difficult to meet the LHC demand, the main problem being represented by conflicting duties (teaching, user support with the older still maintained Fortran package, i.e. Herwig 6.5xx). - HepMC (Outcome of the September MC truth meeting and list of actions agreed by experiments and the Architects Forum): + HepMC should be installed centrally within the LCG external area, as an independent package. + The LCG Generator project should coordinate the requirements from the LHC experiments, the changes to the code and the installation of the package in the external area. + First iteration/verification with M.Dobbs and other people/institutions currently involved in HepMC development/maintenance done. An informal survey with LHC experiments is ongoing. + In the future, the main forum to discuss HepMC, HepPDT developments & maintenance issues would be the LCG-Generators meeting. + The installation in LCG-external area is performed by the LCG librarian. Separation in two libraries: the C++ part; connection with fortran code. - MCDB, HEPML: + Progressing well; now mostly working to document and to populate the data base in order to make it more appealing for the LHC collaborations. We need users to get further feedbacks and collect additional requirements (in particular for what concerns the experiment specific APIs). + The main MCDB developers are giving end-of-the-year MCDB presentations to collaborations meetings (mostly in the context of generator meetings of each LHC collaboration). + Currently verifying possible collaboration with CEDAR on HEPML schema. - Validation: + NRQCD quarkonia production models recently introduced in Pythia 6.3 are now under testing (Marianne). Work plan presented to the October LCG Generator meeting. Progress expected by the end of the year, then Marianne will move to Pythia 8 testing in LHCb. + Interest to start common Validation activities to be developed in the context of the LCG Generator project expressed by many people at the October LCG Generator Meeting. From a practical point of view this is basically already happening. NOTE: Details of GENSER_1_2_0 Release, October 2005 --------------------------------------------- 1. Changes in GENSER configuration: No changes. 2. New versions of installed generators: a) PYTHIA 6.324 b) HERWIG 6.508 (important bug fix concerning top decay) c) LHAPDF 4.1.1 d) ALPGEN 2.0.5 e) photos 2.14, 2.15 (special possibility to switch debugging printout, see README) f) Feynhiggs 2.2.10 g) Sherpa 1.0.6 4. Bug fixes: a) Revision of examples. Added missed call of cernlib in Hijing /examples/BuildFile. b) Definition of PDFsets in Lhapdf is corrected now by making source setPath.csh (.sh). This script sets path to 5. Updates in the examples. 6. Documentation: updated librarian instructions. 7. Tools: a) New scripts to install semiautomatically Lhapdf and Alpgen are prepared. b) Two scripts cmpgen and cmpgenHEAD to compare filesystems of two versions of generator, and given version of generator to the HEAD version of GENSER cvs. 8. Known problems: a) With shared libraries, the library with Block Data should be the first in the list, otherwise it can turn out that Block Data is not loaded. b) With shared libraries, when two libraries with Block Data are used, the case is observed when only part of the second Block Data was loaded. c) On *_dbg platforms on slc3.0.5 the debugger now shows the variables in common blocks, but can show them incorrectly. d) Sherpa-1.0.6 sometimes shows instability in the second authors example. It could be moved to another place in existing file system as a temporary solution. --------------------------------------------------------------------- ==== Fluka - Alfredo No reports. --------------------------------------------------------------------- ==== Simulation framework - Witek - MC-truth: + There has been progress concerning HepMC for the maintenance. The CLHEP version will be moved out. HepMC will be maintained and distributed through LCG within the Generator Services project. At the meeting on September 28th, all four experiments presented their approaches to MC-truth handling. Many commonalities have been noted and three of the LHC experiments (ATLAS, CMS and LHCb) expressed interest in further common work. Technical discussions have been started and first potential implementation aspects have been examined. It became clear that the selection of particles to be stored should be fully configurable with variables like energy, production process, particle types, etc taken into account. Two concrete solutions, LHCb (based on tracking actions), and ATLAS (based on G4VTrajectory::AppendStep method), have been briefly discussed. Further study is planned to examine both solution closely and to understand all the requirements of the general MC truth handling mechanism. Furthermore, it has been noted that some changes on the side of Geant4 could facilitate a lot the MC-truth handling. An example of this could be to associate a numerical (int) identifier to the physics processes allowing for easy checks. Other aspect was the possibility to introduce some sort of 'truth holder' in Geant4 which would be the equivalent of G4VTrajectory (or G4TrajectoryContainer) optimized for the MC-truth handling. Finally, it was mentioned that some sort of 'prioritization' of user actions (or in other words introduction of 'special actions') could be needed in order to avoid conflicts between the 'mandatory' actions (like MC-truth handling) and other, detector specific actions. - Flugg/TileCal: + The goal is to complete the work by end of the year; some progress has been achieved; for the moment focussing on technical issues. GDML is used for exchanging the geometry description; the usage of replicas in the TileCal required implementing some extensions to GDML for the writer, GDML was extended to support this. Hits are produced in Fluka and then later on translated into 'Geant4/ATLAS hits' through the sensitive detector of the TileCal setup, where readout geometry is also used. - GDML: + GDML extended to support and export replicas to writer. + A candidate technical student for a defined project with GDML has been identified for the coming board. + There have been reports from SLAC about GDML at the Geant4 Users Workshop in Bordeaux, the American Linear Collider community is happily using it and they have shown they can easily configure different geometry setups. + An abstract for a presentation at CHEP06 has been submitted. --------------------------------------------------------------------- ==== Geant4 - John - Geant4 Users and Collaboration Workshop in Bordeaux (3-10 November): 120 people were registered at the conference. + Users workshop: mostly focussed to the medical domain, including imaging (PET/SPECT) and hadron-therapy. Key considerations included the need for energy deposition accuracy at or below 1mm for hadron therapy. A small number of HEP related presentations also were given, including one on using LHC-at-home infrastructure for Geant4 regression testing. Detailed tests of physics processes were presented. These included comparisons of proton/hadron stopping in water phantoms and comparisons of Bremstrahlung production in different materials. Different comparisons of proton depth-dose measurements showed the current status, with the beam energy calibration of equipment being a key challenge in benchmarking the MC models. For the newer field of light ion hadrontherapy, a comparison showed reasonable agreement. The first results from an in-depth tests of Bremsstrahlung production of electrons of energy O(10 MeV) were reported. A dedicated experimental setup was built about a decade ago for testing this, with careful measurement of beam energy and a purpose-built detector with low-angular acceptance to measure gammas produced. Results showed good agreement with experiment and speed for heavy materials of important targets, while finding O(10%) differences and slower speed for light materials, which fortunately are less important for the applications. These speed comparisons neglected Bremstrahlung splitting (creating several secondary at each Brem creating interaction, with appropriate weights), which is required and will be introduced in near future. The level of expertise in the application domains and innovative use of Geant4 were impressive. Several users demonstrated the abilities to utilise and customise Geant4, extending the toolkit for the unusual requirements of their application. These included inputing CT scan data from different setups and tayloring the existing navigation or creating customised revised navigation for best speedup; simulating a breathing body with a moving geometry (in several snapshots) and depositing energy in the moving 'voxels' using the ability to change geometries; simulating the beam delivery systems for clinical accelerators, in many cases with several elements customised for each patient; and enabling the use of multiple densities for a single material to simulate 2600 virtual materials using 260 Geant4 materials. + The meeting included also a short course on Geant4, with almost full attendance for approximately 85 users at the Conference. In small sessions in parallel, there were discussions of detailed technical matter including optimisations for particular geometries. Some of the users requested and received invitations to the Collaboration workshop for in-depth discussions. + Collaboration workshop: was held with good spirit; many discussions; dedicated sessions on physics lists. Key aspects included reports from users in established and new application areas, overview of verification/validation, next steps in the evolution of the 'suggested' physics lists. There were updates on the validations and use of Geant4 in HEP experiments and a report from the Spenvis/Geant4 Space Users meeting (Leuven, Belgium October 2005). Summaries were presented of the extensive verification testing of physical models in EM and hadronic physics. Improvements were presented in multiple scattering addressing an issue identified in setups with materials of much-different Z values and densities. The sensitivity of the energy depositions to changes in cut, which were pronounced for particular cases (Linear collider calorimeter, LHCb, simple setups of water phantoms with inserted pertubation layer) and observed also in additional cases (simpler setups similar to ATLAS detectors). It was the occasion to assess the status of the hadronic working-group after the departure of J.P.Wellisch, identify areas of responsibility and decide on revised structure and new leadership. The working-group decided for two co-coordinators and an election was taken for these posts. Aatos Heikkinen (Helsinki Institute of Physics - HIP) and Dennis Wright (SLAC) were elected co-coordinators. An invited guest, one of the authors of INCL inter-nuclear cascade model (adopted recently in MPCN) presented their model and its physics results in particular the range of 100 MeV - 3 GeV projectile kinetic energy. The team has close ties with experimental groups undertaking measurements and comparisons with data, and interest in collaboration with Geant4 were noted. Results using a well established parallelisation solutions using the DIANE tool were also presented. + Collaboration Agreement and CB meeting: SLAC, CERN, Lebedev and HIP have signed the new Geant4 Collaboration Agreement (CA), for a total of approximately 13 FTEs. Groups and individual members were asked to identify who will sign the CA on their behalf, and those without automatic representation (2 FTEs) to band together for representation on the future Oversight Board (OB). The Geant4 CB decided to create an interim OB for the first 3 months after the new CA goes into effect, before moving to elect its chair. + Licence: A dicussion of the effect of the candidate Geant4 license (LGPL, EGEE) was held. A straw poll showed that both licenses were deemed acceptable. Yet most developers remained unsure of the full impact or differences of the licenses. + Slides of the talks will be put available soon from the Geant4 web site. - Geant4 is now a virtual organisation in the LCG/EGEE GRID; the committment of sites for upcoming productions is being sought. Resources required were estimated at 200 GHz-weeks during a 10 day period before the release. News (21 Nov): CERN has now pledged to provide 50% of resources required for next month's release and further sites including Imperial College, DESY and GSI have provided concrete pledges. Further commitments are being sought following the interest shown by additional sites to contribute. Production runs on the GRID for next release's validation will start in the week of Nov 28th-Dec 2nd. - Releases: - The testing phase for release 8.0 is started. The new public release of Geant4 is planned for December 16th. - The reference tag of October includes changes in multiple-scattering which improve the stability of energy deposition in sampling calorimeters. ATLAS is now putting it under test, to provide feedback on the performance impact of the modifications which is under further evaluation. Old options (parameters and models) are foreseen to be kept for use in at least transitionally for some LHC/HEP experiments use cases. A presentation is planned on these developments at the next simulation validation meetings. --------------------------------------------------------------------- ==== Garfield - Rob - Currently involved in NA60 activity, work for the publication of a paper is going on. - Getting now involved in the validation activity, in particular in the understanding of the statistical significance of the resolving power of the tests for the hadronic showers (what is the sensitivity of the tests, what do the shifts mean, and which changes in the MC are relevant in the measurements). Some contacts with people in ATLAS about this kind of study. - Garfield: now concentrated only on support. People seems rather happy with the last updated xsections introduced. --------------------------------------------------------------------- ==== AOB Next meeting in January.