Rendering of Large Scale Continuous Terrain Using Mesh Shading Pipeline

Large open terrains require visual simplifications to be able to be rendered in real-time. One possibility is to subdivide the terrain into a quad tree where each node in the tree represents a simpler version of a tile of the terrain. This requires an algorithm that can traverse the quad tree, choose which tiles to render, and to save performance, remove tiles that are not in front of the camera. When the terrain is simplified in this manner, the algorithm must also make sure that the terrain does not feature holes in between the simplified parts. This thesis investigates the differences between the Mesh shading pipeline and the traditional graphics pipeline in two kinds of these terrain algorithms. First a naive algorithm implemented on the GPU is measured in terms of execution time, it is then changed to use a Concurrent Binary Tree for the traversing of the quad tree and measured to see if this data structure could be utilized to improve the algorithm in terms of execution time. The results show that, while the Concurrent Binary Tree yields a theoretical improvement to a Quad Tree based terrain simplification, the practical application is limited. Regarding the choice of rendering pipeline, the Mesh Shading pipeline did not yield any significant improvements and a conclusion could be drawn that, without any advanced culling algorithms, the traditional pipeline performed better overall. This thesis also presents a continuous rendering algorithm, which draws inspiration from two state of the art rendering algorithms