Optimization Techniques for Game Applications

This bachelor thesis addresses the issue of performance optimization in game applications. The theoretical part defines the architecture of game engines, analyses the stages of the rendering pipeline, and classifies primary performance bottlenecks on the CPU, GPU and memory sides. Within the theoretical framework, geometry optimization methods are described in detail, including Level of Detail (LOD) systems, Nanite virtualized geometry, and a wide range of culling algorithms. Furthermore, procedures for the CPU optimization are analysed, particularly the reduction of draw calls through batching and GPU Instancing techniques, along with the application of optimization design patterns Object Pooling, Data Locality, or Dirty Flag and principles of task parallelization using Job Systems. Attention is also given to pixel processing optimization through Z-Prepass techniques, a comparison of Forward and Deferred Shading models, and the use of precomputed lighting (Baking). The practical part focuses on the measurement and analysis of selected methods within the Unreal Engine 5 environment. Using profiling tools, the benefits of GPU Instancing, Occlusion Culling, Detail Culling, and Baking are shown and quantified. The results confirm that the correct implementation of these techniques leads to a significant reduction in frame time and more efficient utilization of hardware resources.