GPU Accelerated Ray-tracing for Simulating Sound Propagation in Water

Abstract: The propagation paths of sound in water can be somewhat complicated due to the fact that the sound speed in water varies with properties such as water temperature and pressure, which has the effect of curving the propagation paths. This thesis shows how sound propagation in water can be simulated using a ray-tracing based approach on a GPU using Nvidia’s OptiX ray-tracing engine. In particular, it investigates how much speed-up can be achieved compared to CPU based implementations and whether the RT cores introduced in Nvidia’s Turing architecture, which provide hardware accelerated ray-tracing, can be used to speed up the computations. The presented GPU implementation is shown to be up to 310 times faster then the CPU based Fortran implementation Bellhop. Although the speed-up is significant, it is hard to say how much speed-up is gained by utilizing the RT cores due to not having anything equivalent to compare the performance to.