Ray Tracing on Modern Graphics Hardware

Bachelor-Thesis

Rasterization based Rendering Algorithms are limited in the level of realizm they can achieve. Raytracing generates beautiful images, but the rendering times can be very long. The raw floating point performance and the high memory bandwidth of modern GPUs suggest this Hardware platform for ray tracing. GPUs without branching in the fragment stage do not perform very well in ray tracing. The latest hardware generation, Nvidia NV40, provides this feature. It is the goal of this thesis to design and implement a ray tracing kernel for the fragment stage of this GPU. A hierarchical data structure like a kd-tree or an octree has to be used for ray shooting acceleration.


1. Selection of a GPU-friendly acceleration data structure (grid, kd-tree or octree)

2. Design of an efficient traversal algorithm

3. Implementation of the complete ray tracing kernel in the fragment stage of the gpu

4. Comparison with other CPU and GPU ray tracing implementations

Advisor(s)

• Dr.-Ing. Manfred Ernst
• Prof. Dr.-Ing. Marc Stamminger