Simulation project leaders meeting Nov 20, 2003 =============================================== === Agenda - around the table on project status and news, including milestone review and manpower status http://atlassw1.phy.bnl.gov/XProject/servlet/XProject?schedule=yes&proj=lcg&keys=simu&brief=yes - GDML proposal - AA internal review - LHCC review next week - developer-experiment associations in the simulation project - simu project presentation plans for upcoming apps area meetings. - Date for next meeting === Attendees Fabiola, Paolo, John, Torre (Alfredo and Andrea were not able to attend) === Minutes - around the table on project status and news, including... == Geant4 Release 6 is in preparation. It is due in three weeks and is expected to be on time. Geant4 has three deliverables to the project in this quarter: - Performance on benchmark suite: should be done by the December break. Not a lot of work. - G4-SPI collaborative pilot: Delayed because Gabriele has been away. The time window before release 6 preparations could not be used. Reset milestone to mid February. - Release 6 should be on time, Dec 12. Will reflect a lot of good feedback from the experiments. John notes that in 2004 the balance of work between new development and experiment support (responding to feedback, supporting production usage of Geant4 in experiment DCs, etc.) will shift towards experiment support. There will still be a development effort in 04 but it will be smaller. The Geant4 project team is developing an update to their workplan now. A prototype abstract interface to the G4Navigator is available now. Whether it will be part of the 6.0 release is not decided yet, but it is available for people to try out. John will investigate whether there might be Spanish interest in contributing to the G4e work. Physics lists will be in release 6. == Physics validation Fabiola confirms the milestones currently in the schedule. The first cycle of em physics validation is complete, meaning that Geant4 is equal to or better than Geant3 in comparison to data. Work still continues however, e.g. looking at tails and lateral shower profiles. This is going on for instance for the ATLAS ecal strips, and more results are hoped for at coming physics validation meetings. In hadronic physics validation, work is progressing well in ATLAS. The main issue now is to check whether the QGSP physics list which looks good for ATLAS is also the best list for the other experiments. There should be CMS HCAL results by the end of the year, maybe in the Dec meeting. LHCb is less clear because nobody is working on it at present. The project may contribute effort to help LHCb check their test beam data against the QGSP list. Juerg has finished his simple benchmark study of double differential cross sections and is finalizing a note. Before he leaves at the end of the year he plans to do a second study, looking at pion absorption. He has done a nice job of setting up all the required code for simple benchmarks in the CVS repository so that it will be relatively easy for others to carry on the work. This work also constitutes a good first step in putting together the validation test suite. Fabiola notes that, particularly now with the code infrastructure that Juerg has set up, contributors can get involved by focusing on a single study with only a fraction of their time; such contributions are welcome. In the absence of experiment participants this work will be high on the priority list for applying LCG project effort. Fabiola's big concern is who will contribute after Juerg's departure. The AA internal review committee made the often heard comment that Physics validation is an ideal area for young physicists to contribute. As the internal review report also said, the experiments require accurate and high performance simulation tools; this project and these studies are important part of delivering that. And yet the experiment manpower working on the simulation physics validation using test-beam data is quite small (except in one experiment). Fabiola is also investigating possible Spanish participation, in particular to the simple benchmark studies. Regarding FLUKA validation, the plan was to launch two projects, in ATLAS tilecal and pixel hadronic interactions. The tilecal work was expected to be underway by now, but it requires infrastructure work by Andrea to interface FLUGG to the simulation framework. This is non trivial because the FLUGG interface has changed in the two years since it was last used by ATLAS. This is essentially a statement that we need the generic simulation framework prototype, cf. the discussion on that below. == Generator services Paolo has just done a careful review and update of the milestones; the revised milestones are on the web and Paolo confirms the dates. The strategy to achieve them is still being developed. The long term goal is ambitious, it is to provide a common production event generation operation for all the experiments to build up a database of standard generator events using generator code coming from a common library of generator packages. It should also include a framework for the comparison of generator results as suggested by the internal review. The agreed tool for event database storage and bookkeeping is MCDB, the official version of which is in GENSER. It is being used now by CMS and at Fermilab. There will be a paper by the end of the year with a proposal for MCDB deployment by the project. Lev Dudco the main author of MCDB is contributing to this. For production, the plan is to convince the experiments to adopt soon the GENSER library and integrate its use into their frameworks and generator-level production environments; the project can then evaluate these experiment environments for use by the project. The project cannot develop a new production infrastructure from scratch. Using experiment infrastructure, with the possible need to modify storage and event interfaces to the standards that are agreed, will be a minimum-manpower way to put in place the capability to build up a common generator event database. Paolo reports that GENSER is being positively received so far, with work underway in some experiments to implement GENSER in their simulation frameworks. Paolo is also encouraging the experiments to evaluate MCDB for storage and bookkeeping of partonic level events. At experiment request the LHAPDF generator has recently been added to GENSER. A new GENSER release is planned by the end of the year. The last item in the chain is is the validation of the generators. But at the request of the experiments some validation work has begun already, in particular with HIJING, the generator most recently added to GENSER. Vladimir Oujinski is working on this and has uncovered significant bugs that require a local bugfix version of HIJING to be put in place because the turnaround time to get a fixed version from the authors will be too long. A paper on the HIJING work is being prepared. Paolo notes that there are deeply modified versions of HIJING in the experiments and it will be good to have an official CERN version. In the generator validation area there is clearly an interplay with the physics validation activity in this area. The tools MC-Tester and JetWeb will be used for generator comparison and validation. Paolo is investigating possible Spanish interest in participating in common generator event production. Paolo is also beginning to plan the computing requirements for a common production operation. == generic framework Andrea was not able to attend; Torre had a discussion with him recently. Most important for the generic framework activity is to meet the end-year milestone for the delivery of a first prototype. As settled in a recent AF, what this prototype will be is simply an infrastructure supporting the use of FLUKA as well as Geant4 as a simulation engine via the FLUGG FLUKA-G4 geometry interface. As noted above, this is needed immediately by the physics validation project such that test beam simulations with a Geant4 geometry can use both Geant4 and FLUKA and so allow the FLUKA validation to go ahead. Andrea expects to be able to concentrate on this work in December, with help from Witek. If he can do this he expects to meet the milestone or come close (early Jan). An issue for a long time has been the adequacy of the handling of MC truth information in the simulations and particularly the support for navigating up the particle stack from secondaries to parent generator-level particles. For example there is this statement in the last SC2 response to PEB quarterly report... The consistent support and treatment of MCtruth information (together with HEPMC) must be taken into account when developing the Generic Simulation Framework and for Fluka and G4. It is suggested that progress in this area be reported through the Generic Simulation Framework part of the project. Resolving the issues linked to MCtruth in general, HEPMC in particular, it is suggested to have a dedicated review process and a new high-level milestone. ... it has been deeply confusing at least to the minute writer what the status really is and what exactly the complaints are directed at. Within this meeting the belief is - HEPMC has had serious maintenance and responsiveness problems which in the past has led to doubts over whether it is a viable final solution. The situation is better now however; not all is resolved but progress is being made and HEPMC is accepted. - Navigation in the particle stack is supported adequately for MC truth matching in physics studies. It works, it is stable, though it may not be optimal and could be improved. - persistent storage of generator data has been an open issue and is now being addressed in the generator services project, cf. their last meeting and the next one later this month. == GDML proposal Torre and Andrea will prepare a proposal to bring GDML into project scope following the plan of the recent AF meeting. Andrea will present this to the SC2 in December. T&A will also develop an updated set of milestones for the generic framework subproject following the plan of the recent AF http://lcgapp.cern.ch/project/mgmt/AF20031009.txt == AA internal review Comments on the internal reviewer's comments on the simulation project are requested. Torre will prepare a written response to the review report comments which will be part of an overall AA response. == LHCC review next week In the next day or two Torre will circulate a draft of the simulation talk for the LHCC review of LCG next week. == Developer-experiment associations in the simulation project Simulation subproject leaders are asked to comment to Torre on this soon. == simu project presentation plans for upcoming apps area meetings. The following are agreed: - Fabiola will invite Juerg to give a talk on simple benchmarks Dec 17 - em physics validation and physics reqmts review talks will be scheduled circa end Jan, Feb - A talk on GDML and its role in the generic framework plan will be planned for ~late Jan, after SC2 proposal presentation and generic framework 2004 planning - Paolo can give a talk on the GENSER new release and other generator services news at the first meeting in Jan - Geant4 release 6 talk early in the new year == Date for next meeting Tentative date is Jan 14, 10:00.