Kiran Aanandi

Advisor: Hector de la Torre Perez

Mentor: Walter Hopkins

Undergraduate: (Physics and Mathematics (minor in French))

Graduate: Northern Illinois University (Physics)

Project: Celeritas-Athena Integration

Graphics Processing Units (GPUs) have emerged as powerful computing resources, offering improved power efficiency compared to traditional CPUs and increasingly shaping the landscape of high performance computing. Despite this shift, the Geant4-based ATLAS detector simulation, known as FullSim, still operates exclusively on CPUs, limiting its ability to benefit from GPU acceleration as these architectures become more prevalent. Recent progress has been made through the development of Celeritas, a framework designed to enable detector simulations on GPUs, though it has not yet been fully integrated into the ATLAS FullSim workflow. This work focuses on advancing that integration through a series of targeted studies. The first objective is to execute the full validation chain both with and without Celeritas, documenting the specific simulation flags required to successfully run with GPU support. The second objective examines the impact of selectively disabling optimizations, such as Neutron Russian Roulette, in order to understand their influence on performance and to identify the most effective GPU configuration. Finally, the study compares the computational performance of an optimized CPU-based setup against an optimized GPU-based configuration, providing a clear assessment of the benefits and tradeoffs associated with adopting GPU acceleration in the FullSim framework.