The AutoPas particle simulation library uses automatic tuning to optimize node level performance. This work implements short range particle interactions for molecular dynamics simulations on GPUs for all containers currently implemented in AutoPas. The basics of molecular dynamics and the CUDA programming language are presented and the implemented methods are introduced. Benchmarks were performed and the results for multiple domains are presented. The scaling on GPUs is discussed and the GPU implementations are compared with each other and the current implementation on CPUs.      «

The AutoPas particle simulation library uses automatic tuning to optimize node level performance. This work implements short range particle interactions for molecular dynamics simulations on GPUs for all containers currently implemented in AutoPas. The basics of molecular dynamics and the CUDA programming language are presented and the implemented methods are introduced. Benchmarks were performed and the results for multiple domains are presented. The scaling on GPUs is discussed and the GPU imp...     »