My pre-scientific thesis (VWA), submitted in 2023.
Abstract #
N-dimensional rendering is a fascinating but computationally expensive process. This makes it ill-suited for real-time applications. GPUs have been conventionally used to accelerate rendering workloads, however they are heavily optimized for 3D rendering. This thesis explains the mathematics underpinning n-dimensional rendering and takes a look at how it can be optimally realized on modern graphics hardware. It will give some workarounds to overcome the inherent limitations of components designed not with this use case in mind.
A reference implementation written in C++ and utilizing Vulkan is provided in the form of vulkan-xd, exemplifying some of the techniques discussed. Some basic performance analysis was conducted on the project giving a benchmark for the achievable efficiency.
Further research is required to come up with additional optimizations and to accurately assess the exact performance to be expected.