high performance computing on graphics processing units: hgpu.org

One of the most time-consuming parts of the simulation of underwater networks is the realistic simulation of underwater sound propagation. Some well-known software tools used for networks simulations to date employ ray tracing to simulate sound propagation. This gives rise to high computational complexity, and may require very long time to complete a simulation. In this thesis we present a faster tool able to simulate an underwater sound channel. The software is based on ray tracing and is based on the CUDA architecture, the general-purpose parallel computing architecture that uses the parallel computing engine in NVIDIA GPUs to efficiently solve complex computational problems. Following the CUDA programming model guideline, a ray tracer has been developed where the trajectory computation is performed in a CUDA-enabled parallel way, obtaining a ray tracer implementation that completes faster than the widely use single-thread Bellhop software.