An efficient mulithreading algorithm for the simulation of global illumination

Abstract

The calculation of global illumination taking into account not only the direct light but also all its possible reflections is an important part of the synthesis of realistic images. The simulation of global illumination requires a lot of computational resources. The performance of modern computers directly depends on the number of cores, which allow one to use tens of parallel computation threads. The paper is devoted to the effective use of multicore computers for simulating the global illumination. The computation results are represented in a uniform illumination map. The proposed algorithm dynamically creates additional threads for the critical branches of the computation process or processing of results. As a result, the uniform load of computation threads as achieved, and the physical and virtual cores are fully loaded.