Parallel three-dimensional numerical simulation of rotating detonation engine on graphics processing units

• Detailed descriptions are provided about CUDA optimization on different data structures.
• Speed-up performances vary on different data structures or different GPU devices.
• Memory management and manipulation are the main factors of speed-up.
• MPI + CUDA are an efficient way for three-dimensional parallel simulation on detonation.

Rotating detonation engine is a new concept engine driven by detonation combustion for aerospace vehicles. In this paper, three-dimensional numerical simulation was used to investigate the combustion process and kinetic properties of reactive flow within the combustion chamber. To utilize fine grid in simulation, it requires a significant amount of GPU Time to depict this combustion. With the help of NVIDIA’s CUDA, the computation was discretized and distributed to multiple graphics processing units, making the computation time reasonable and tolerable. The present study showed both the way to map our model into CUDA programming and the performance of acceleration, explicating the details of CUDA programming for different types of data. The result verified that CUDA was an efficient method for detonation simulation.