Research And Implementation Based On Multi-core PC Cluster Parallel Rendering System

With the rapid development of computer hardware, the application of multi-core platforms are more and more widely, but requirements of people on the computer are getting higher and higher, the increasing scale of scientific computing. To meet the requirements of the large-scale virtual reality application in rendering speed and displaying resolution, a high-cost-effective distributed graphics rendering system, being based on a cluster of personal computers with multi cores, was built. Distributed rendering system and multi-core platforms are integrated, and the efficiency of parallel rendering systems was improved by multi-threaded parallel optimization. PC cluster parallel rendering system optimized in multi-core platforms had a higher cost-effectiveness and better scalability.This paper fully integrated the parallel in a node of the multi-core PC cluster and the parallel inter nodes. Being based on a cluster of personal computers with multi cores was designed and implemented which is a large-screen projection and stereo display system. System achieved a coarse-grained distribution parallel in a node of the multi-core PC cluster and a fine-grained multi-threaded parallel in the parallel inter nodes. Main contents of this paper are as follows:(1) This paper gives some research on the basic principles of scene graph and rendering process with the OpenSceneGraph. It studies the basic principles of common scene graph and characteristics, and analyses rendering processes based on a typical pipeline different from rendering processes based on the scene graph. It analyzes the rendering background panorama of OpenSceneGraph, several important stage of rendering pipeline in the OpenSceneGraph, and the factors that affect the rendering which realized the basis for distributed rendering and the multi-threaded parallel optimization.(2) A sort-first distributed rendering system based on scene graph is researched and implemented. Through the analysis of system requirements, a common framework of distributed rendering system of more detailed parallel granularity is summed up. It analyzes the system's process of concrete realization and especially has a deep research on the classification of the distributed rendering system, frame synchronization mechanism, seamless splicing and edge integration, network mode and so on. A more flexible approach of Split-screen based on OSG is achieved with scaling the frustum and moving the center of the projection.(3) Multi-core parallel optimization in the parallel inter nodes is studied and implemented. This paper studied methods of multi-threaded parallel optimization and multi-threaded programming problem that needs attention on multi-core PC. Through an analysis of the parallel region on the multi-core platforms, the systems is optimized by multi-threaded parallel optimization of the rendering pipeline, loop iteration, function-level in the cluster node.Experimental results show that:parallel rendering system based on multi-core PC cluster not only greatly improved rendering speed, but also achieved large-screen projection and stereo display. System fully utilized the hardware resources and provided a solution for rendering large-scale application. The system of combination of multi-core platforms and distributed rendering system has more cost-effective and better scalability.