Simulation Project Leaders Meeting 9-July-2013 ==== Attendees (virtual meeting) John Apostolakis (JA), Alfredo Ferrari (AF), Witold Pokorski (WP), Rob Veenhof (RV), Gabriele Cosmo (GC). --------------------------------------------------------------------- ==== Minutes --------------------------------------------------------------------- ==== Simulation physics validation & framework - Witek, Alberto - Activities: o First validation studies made on release 10.0-beta of Geant4 in multi-threaded mode reveal excellent statistical agreement in the physics observables for the MT-migrated simplified calorimeter test. Event reproducibility has been measured in 91% of the samples tested, which improved to 99.9% (1 failure per mille in MT-mode) after a fix introduced in the current development version. Aiming to provide perfect agreement versus sequential runs for the final release in December. - Meetings: o After the first contacts with experiments representatives for the organisation of the second LPCC workshop on physics validation, a suitable date to hold it appears to be sometime in the beginning of 2014. --------------------------------------------------------------------- ==== Generator services - Witek - GENSER: o The migration to the new system based on CMake has been completed. All the generators required by the experiments are now installed automatically together with LCG externals. Releases of the Monte Carlo Generators for LCG64 and LCG65 have been made. The GENSER team has started working now on the migration of the tests for the generators to CTest. The next release based on LCG66 will be done centrally, by the LCG externals librarian. The installation of the generators on AFS and CVMFS will happen together with the rest of the LCG externals. --------------------------------------------------------------------- ==== Fluka - Alfredo - No report. --------------------------------------------------------------------- ==== Geant4 - John, Gabriele - 10.0-beta release: o The Geant4 10.0-beta release was made on June 28th. It incorporates the multi-threading capabilities of Geant4 MT. At build time, it is possible to choose whether to install the sequential version or the multi-threading capable version of Geant4. Several examples have been migrated to make use of multi-threading. o The old GNUMake build system is no longer supported. It is recommended to adopt CMake for the installation. o The GeomTest check for overlaps is withdrawn; the functionality of overlaps checking is available using random points located in surface. Removed deprecated BREPs module and obsolete scorer classes. Improved isotropic safety in the geometry modeller for voxelised volumes is now the default. Field propagation has been refined to allow simultaneous use of gravity field and magnetic field gradient. An existing issue for exit normals in case of multiple geometries has been corrected. o NIST element and material builders always create elements with natural abundances of isotopes. o The QGSP_BERT physics-list has been revised, completely replacing the use of parameterised LEP models with Bertini and FTFP. For nucleons and pions this affects the energy range 9.5-25 GeV. Removed CHIPS models and the physics lists and configurations that used it. Major restructuring in the physics list module, refactored in new submodules. o In Bertini introduced improved multi-body final state momentum distributions, to address deficiencies above 4 GeV. Corrected the sampling of internal nucleus interaction point for gammas. Added detailed two-body final state angular distributions for (gamma,p) and (gamma,n) reactions, and also for nucleon-nucleon scattering. Including isomer nuclei systematically for the first time. First treatment in de-excitation module. o Providing new version of the Urban multiple scattering (G4UrbanMscModel). Modified sampling of cosine theta to provide higher stability versus step size. Removed obsolete versions ('tunes') G4UrbanMscModel90 and G4UrbanMscModel92. New G4UrbanMscModel is the default model for muon multiple scattering (however production physics lists QGSP_BERT and FTFP_BERT still use the Wentzel model). o New data sets are provided and required: G4EMLOW-6.33 for EM processes including Bremmstrahlung and Livermore model, G4NDL-4.3 and G4NEUTRONXS-1.3 for the detailed data-driven (HP) and simpler improved cross-sections (XS) of neutrons and G4RadioactiveDecay-3.7. o Migration issues: headers of Geant4 classes no longer include SystemOfUnits and PhysicalConstants headers from CLHEP. It is expected to trigger compilation errors in user code, requiring to include such headers explicitly. o Code revision and improvement: many classes were reviewed for multi-threading, and a significant number of revisions were made to ensure thread-safety and simplify the code from the first implementations from the Geant4-MT prototype. A key example is the revision of physics vector classes (1d and 2d) to be read-only; a cache is no longer provided. Client classes in the toolkit which are performance sensitive were changed to implement their own caching. Other changes enable memory management in the multi-threaded mode: G4Allocator has been revised to enable deletion of allocator objects in the case of multi-threaded builds. Further revisions will be required to manage objects which are used to be statically created in class methods, but are now created on the heap. - Development: o First performance measurements on 10.0-beta release reveal excellent linear scaling versus the number of worker threads, in the order of O(100) threads measured on the Intel Xeon-Phi co-processor, and good save of memory (40 MB/thread for a benchmark based on full-CMS geometry). Aiming to reduce memory usage even further for the final release. - Experiments feedback: o Validation of release 9.6.p02 has been completed in CMS and is now planned for use in summer production. - Meetings: o 18th Geant4 Collaboration Meeting in Sevilla (Spain) on September 23-27th. o Geant4 User Conference on Medicine and Biology applications in Bordeaux (France) on October 7-9th. --------------------------------------------------------------------- ==== Garfield - Rob - Development: o Ion mobility and diffusion: contact established with a group in Coimbra to discuss in detail measurements to be made. Will start in August when the new student who is going to work on this, has arrived. o Transport: measurements for Ar-CO2 have been completed, scanning both the pressure and the mixing proportions; measurements of Ne-CO2 are nearly complete; Xe-TMA is underway. o Summer students are working on a variety of topics, e.g. a way to visualise transport parameters with varying mixing proportions; limiting ion back flow and protecting cathodes against ion bombardment. - Meetings: o MPGD 2013 and RD51 collaboration meeting held in Zaragoza. Particularly well organised. --------------------------------------------------------------------- ==== AOB Next meeting scheduled for October 29th.