high performance computing on graphics processing units: hgpu.org

During the past two decades, the Dutch National Aerospace Laboratory has developed a model to calculate the risk for third parties around airports. This Third Party Risk model is used in the decision making with respect to airport development and land use planning. Due to the increase of air traffic, the availability of improved individual flight track data, and the need for detailed calculations on a denser grid, the time required for a risk calculation has increased significantly. In this thesis, we present the research of using parallel programming hardware, in particular Graphical Processing Units, to accelerate a risk calculation. Because of different phases in the calculations, and the relation between adjacent grid cells, the translation of the Third Party Risk model into a parallel implementation is not straightforward. Results show that a calculation for Schiphol Airport, based on the traffic for a calendar year and a dense grid that took a week to complete in the original implementation, finishes well within fifteen minutes in the improved parallel implementation.