Correction Based On The Machine's High-performance Pc Cluster Parallel Graphics Rendering And High-resolution Projection Wall

With the the rapid expansion of information today, people are becoming more and more interested in the rendering speed, display size, resolution of 3D graphics rendering. The common solution to this problem is just to use multi-processor, multi-pipelined graphics workstations. But its high price greatly limits its popularization. Nowadays, with the great improvement of PC-level hardware, the performance of low-level graphics cards double nearly every 6 months. So, in our system, we use PC cluster to replace the big iron of SGI workstations, use distributed parallel rendering to replace single chip rendering, and thus realize the real time rendering of large scale scene data. On the other side, considering the eager for super resolution display system and the limitation of single body display, we utilize multi-projector to realize a real super resolution display system.In this paper, we will introduce in detail our 32-node PC cluster for high performance graphics rendering, the 30-projector tiled high resolution display wall, and the two sets of software system on this hardware platform, i.e., retained mode based distributed parallel rendering, digital camera based display wall calibration.First, we will address the problems encountered in the construction of our PC cluster, i.e., hardware selection, high performance communication networks, and the problem of remote control of our cluster system.Second, we will present our retained mode based distributed parallel rendering system-MSPR. The chapter includes system architecture, implementation of clients and servers, data distribution and management, and results comparison.Third, we will introduce our 30-projector tiled high resolution display wall, and address the problems of projector and screen selection, stereoscopic display implementation, etc..Then, we will present our software calibration system based on a digital camera, and explore the background, algorithm introduction, and results analysis.Finally, we will come to the conclusion and explore the possibility of further work.