Advancing Efficiency of Unstructured Mesh Processing With Localized Data Structures

Unstructured meshes are widely used to represent complex shapes and data in visualization tasks, such as medical imaging, engineering design, and geometric modeling. However, their unevenly distributed elements make them memory-intensive and time-consuming to process, especially as mesh sizes grow. This research focuses on improving the efficiency of processing large unstructured meshes by reducing memory usage and speeding up computations. This doctoral dissertation introduces three methods to address these challenges. The first approach divides the mesh into smaller partitions and processes it piece-by-piece, reducing memory requirements by up to 10 times. The second approach uses the processor's parallel computing capabilities to accelerate the processing speed, achieving up to 3 times faster performance while keeping low memory usage. The third approach combines the power of both the processor and the graphics card to further improve performance by offloading some computational tasks to the graphics card, resulting in nearly 3 times faster than previous methods. Together, these advancements make it easier and faster to work with large unstructured datasets in fields like medicine, engineering, geography, and material science. By improving the efficiency of unstructured mesh processing, this research enables scientists and engineers to analyze data more effectively, leading to better insights and discoveries.