Simulating Collective Effects on GPUs

Computer simulations are an important tool to study the dynamics of charged particles in particle accelerators, with new hardware solutions such as GPUs providing a vast increase in computing power. In the accelerator physics domain simulations are used to understand instabilities arising due to collective e↵ects in high intensity beams which limit the accelerator performance. In this thesis PyHEADTAIL, a code to study collective effects in synchrotrons, is ported to GPUs using PyCUDA. The goal is to achieve a significant speedup while at the same time producing a simple interface for users and other developers. A speedup of 6 compared to the CPU version is achieved on a typical simulation study of instabilities in the Large Hadron Collider (LHC) at CERN.