An alternative e-/e+ and gamma transport simulation highly specialized for HEP detector simulations
R&D task number: G4RD5
An alternative e-/e+ and gamma transport simulation highly specialized for HEP detector simulations
The current simulation framework used in Geant4 (process interface, track, stepping algorithm, etc.) is completely general. From one side this generality provides a high level of flexibility that can serve simulations with very different characteristics. On the other hand, it brings many potential sources of inefficiency in case of individual, well defined simulation problems such as HEP detector simulations where only a small fraction of these functionalities are actually required.
This is true even more if one focuses only on those particles and their interactions that are the major consumers of the resources in a HEP detector simulation like electron and gamma particles with their electromagnetic (EM) interactions.
The goal of this task is to investigate the possible performance benefits of a new, simplified physics simulation “framework”, tailored specifically for HEP detector simulations, targeting only the performance critical particles and their interactions with the highest level of specialization for HEP usage. Significant performance benefits can be expected over the current general version due to the highly reduced complexity and size of the corresponding code, more cache efficient data storage designed by considering the run time access pattern, etc… While targeting the highest level of specializations and code simplifications, the goal is to keep the current physics simulation performance without any compromises.
Upon success, merging the result of this development with the current, more general Geant4 framework (that will be our target right from the beginning) would provide immediate performance benefits for all HEP detector simulation users. Moreover, taking into account that this activity isolates, separates and simplifies to the highest level of the most performance critical part of HEP detector simulations, the result can also provide a perfect starting point for further R&D activities targeting GPU implementation of the HEP detector simulation.
Lead and main developers: Mihaly Novak (mihaly.novak@cern.ch)