The Research Of Architecture And Very Important Technologies For Parallel Graphics Rendering System

Along with the improvement of theories and technologies of computer graphics, the technology of graphics rendering is more and more widely used, and has become an important part of human life. As computer graphics technology become more practical, it need to construct large scale data scene more reality and refined. But only depending on the technology of hardware rendering, the need of practical application could not meet. And, only rendering graphics on one computer, there are not enough resources for computing and graphics rendering. Therefore, it is low and can not meet the needs too. So, introducing theories and methods of parallel computing in the field of graphics processing, called parallel rendering, became an important technology to implement high performance graphics rendering.As the development of network technology and graphic cards costing lower in PCs, there is a new parallel rendering system which builds on PCs connected by high-performance networks or workstation clusters. It has the advantage of low cost, preferable expansibility and wonderful agility, and enables creation of powerful parallel graphics applications capable of fully exploiting the power of clusters. So, researching the technologies of parallel rendering on cluster, and designing the related speediness algorithm of parallel rendering is a certain trend and is very necessary.Researching the architecture of parallel rendering system is the kernel of all parallel rendering technologies. Other very important technologies include tasks partitioning, assigning, and load-balancing. This thesis made very deeply researching of the architecture and important technologies for parallel rendering. The main work and acquired production are listed as following:1. This thesis puts forward a new highly parallel architecture of multi-task parallel graphics rendering system for large scale data scene. It supports parallel processing of multi-task and multi-screen display mode. This system separates tasks of geometry computing from graphics rendering, and parallelizes them. It has wonderful parallel efficiency and expansibility.2. This thesis brings forward a model of two phase mapping of dynamic load-balancing algorithm, and gives a method for measuring its performance. One of the best assignment algorithms for multi-task is given. The performance of this algorithm is analyzed, and a result of an upper limited time of parallel rendering is acquired in theoretically. Then we analyze the strategy of partitioning multitask, and give some methods of partitioning. This algorithm has the best assignment scheme for multitask under given granularity of task, and achieves wonderful load balancing effect for suitable partitioning strategy of multitask. 3. Applying parts of the researching production, a simple prototype for parallel rendering system is implemented. It implements dynamic assignment for tasks of graphic rendering and parallel compositing. At last, the primary work is summarized, and some future works are stated.