Performance profiling for brachytherapy applications

Published: 28 February 2018| Version 1 | DOI: 10.17632/kv3rs8dxhp.1
Contributors:
Wonqook Choi,
Kihyeon Cho,
Insung Yeo

Description

In many physics applications, a significant amount of software (e.g. R, ROOT and Geant4) is developed on novel computing architectures, and much effort is expended to ensure the software is efficient in terms of central processing unit (CPU) time and memory usage. Profiling tools are used during the evaluation process to evaluate the efficiency; however, few such tools are able to accommodate low-energy physics regions. To address this limitation, we developed a low-energy physics profiling system in Geant4 to profile the CPU time and memory of software applications in brachytherapy applications. This paper describes and evaluates specific models that are applied to brachytherapy applications in Geant4, such as QGSP_BIC_LIV, QGSP_BIC_EMZ, and QGSP_BIC_EMY. The physics range in this tool allows it to be used to generate low energy profiles in brachytherapy applications. This was a limitation in previous studies, which caused us to develop a new profiling tool that supports profiling in the MeV range, in contrast to the TeV range that is supported by existing high-energy profiling tools. In order to easily compare the profiling results between low-energy and high-energy modes, we employed the same software architecture as that in the SimpliCarlo tool developed at the Fermilab National Accelerator Laboratory (FNAL) for the Large Hadron Collider (LHC). The results show that the newly developed profiling system for low-energy physics (less than MeV) complements the current profiling system used for high-energy physics (greater than TeV) applications.

Files