Simulation Project Leaders Meeting 29-October-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 Identified source of changes in hadronic showers, including energy response and lateral shape, observed in recent development releases. Found that it is due to refinements and tuning made to FTF and Bertini models. - Meetings: o The 2nd LPCC Simulation Workshop will take place at CERN on March 18-19th. --------------------------------------------------------------------- ==== Generator services - Witek - Releases: o New MC generators tress LCG-65a and LCG-66 has been released in August and last September respectively. New version of packages incliuded are: crmc 1.0, pythia8 (176, 175.lhetau -atlas patch- and 180), pythia6 6-428, herwig++ 2.7.0 + thepeg 1.9.0, rivet 2.0.0 ( + yoda 1.0.4 + cython 0.19.1), tauola++ 1.1.3, sherpa 2.0.0, lhapdf (5.9.1, 6.0.2 and 6.0.3) o New generators, evaluated and studied for integration but not really used or accepted: jhu 3.1.8, madgraph5 (not yet added to any release). - Testing: o Implemented "sanity" tests for all main generators, switched from building scripts to the Electric Commander tool; now having daily nightly builds. o Almost all generators can now be built on MacOS platform. --------------------------------------------------------------------- ==== Fluka - Alfredo - No report. --------------------------------------------------------------------- ==== Geant4 - John, Gabriele - Development: o In preparation for release 10.0 of December, there has been considerable developments in the physics: latest developments to FTF have been incorporated; improved electro- and photo- nuclear cross-section, to address a performance issue reported by ATLAS (cross-section now depends only on Z, and utilizes the average number of neutrons N; replaced costly 'per isotope' methods by 'per element' cross-section). Further adaptations for isomers to use an integer order to identify the metastable-state level instead of an arbitrary energy values; where only energy is known in the method for nuclear decay channel, search for the nearest (integer) level. Unified naming convention of isomers and general excited nuclides; they always appear with excitation energy in unit of keV. Reviewed the PAI model implementation, and migrated it to multi-threading. Added isomer production into neutron capture model. Added ESTAR stopping power data in EMLOW data-set and corresponding class to use it. Using Bertini cascade instead of pre-compound model for muon capture. o Fix to avoid second initialisation in Muon Bremsstrahlung model (now shared with anti-muon). o Fixes made to G4OpBoundaryProcess to work correctly with parallel worlds. - Computing Performance: o Introduced fast log and exponential functions, extracted from VDT, in key areas of 'Standard' EM physics models, Bertini cascade, pre-compound and evaporation module. These replace standard library functions: exp log, log10 and pow (power). A reduction of CPU time for electron/gamma simulation is foreseen at the few percent level. o Revised physics-vector classes, eliminating virtual function calls and using the fast logarithm function from VDT. o Reduced memory churn in INCL cascade. - Event reproducibility: o Extended reproducibility testing for multithreaded runs. Improved reproducibility of multi-threaded versus sequential Geant4 runs; achieved perfect reproducibility for most physics configurations tested. o Extended routine testing to data-driven low-energy neutron modeling (HP), radioactive decay and INCL models. o Identified and corrected reproducibility errors in revision of muon physics model, in radioactive decay and in the initialisation of the table of isomers. o Currently the only outstanding issue is in multi-threaded mode when using the INCLXX models. - Multithreading: o New ability to specify the number of events to be sent to a worker thread in one go (it used to always be one event). o Introduced flag shell variable to override the number of threads set by the G4MTRunManager or by the corresponding UI command. o Revised split class for G4ParticleDefinition so that all generic ions share the object ID with GenericIon as a result one G4ProcessManager is used in each thread for all ions. o Check for support of thread-local storage in cmake configuration. o Improved Workspaces in Geometry (for extension to TBB) and enabled its use in MT mode. o First version of new G4Cache classes for handling thread-private data members in shared objects. Added new lock protection for registration and de-registration of worker threads and re-initialisation of geometry. o A UI command directory can be set so all commands are not broadcast to worker threads. This can be overridden on an individual command basis. - Meetings: o Geant4 Collaboration meeting has been held in Seville (Spain) last September. Main focus on reviewing the developments ongoing for release 10 and progress made in the different areas. The whole program can be found online on Indico: https://indico.cern.ch/conferenceDisplay.py?confId=250021 There have been also reports on GPU prototypes being developed by various teams at FNAL, KEK/SLAC and OpenGATE. o Next Geant4 Collaboration meeting will be held in Okinawa (Japan) in fall 2014. --------------------------------------------------------------------- ==== Garfield - Rob - No report. --------------------------------------------------------------------- ==== AOB Next meeting scheduled for February 18th.