CPU-GPU Parallelism Applied to Quality-Diversity Volume Rendering
Cristian Nicolae Buţincu, Marius Gavrilescu · 2024
This paper presents the performance analysis and the most significant design decisions taken to improve the runtime performance of a quality-diversity volume rendering framework, focusing on using parallel and distributed computing to optimize specific stages within the framework pipeline. After considering the benefits and trade-offs of several aspects like minimization of execution time, portability and hardware independence, we developed a hybrid CPU-GPU parallel computing model that increased the overall efficiency. Using this model, we achieved significant improvement in runtime speed for critical components of the framework compared to the sequential implementation. Additionally, this model can be applied in a wide range of scenarios, where the performance improvements provided by GPU processing can be complemented by CPU parallelization, with better runtime results.