|Name:||(PP22) GeantV: Designing the Future of Particle Transport Simulation for HEP|
|Time:||Tuesday, June 20, 2017
03:15 pm - 03:45 pm
|Breaks:||03:15 pm - 03:45 pm Coffee Break|
|Presenter:||Sofia Vallecorsa, CERN|
|Abstract:||Approximately half of today’s LHC Grid computing power budget is used for detector simulation. The current code, GEANT4, shows suboptimal performance on modern SIMD architectures. A new version, called GEANTV, is being developed to improve the physics accuracy, while preserving, at the same, portability through different architectures (Xeon Phi, GPU). Introducing fine grained parallelism, vectorisation, efficient memory management and NUMA awareness, we achieve considerable speedup with respect to current GEANT4 and a good scalability through a large number of threads. Furthermore, we will present preliminary results on a new machine learning based approach for fast detector simulation. This new development, within the GEANTV framework, is intended to address the ever increasing need for simulated events expected for LHC experiments and their upgrades, such as the High Luminosity LHC.
Author:Sofia Vallecorsa, CERN
PP22_Vallecorsa.pdf (17109 KB)