InterConnection Network And Memory Design For Parallel Volume Rendering Architecture

Ray Casting is one of the image order algorithms for the volume rendering problem. it can produce high quality image but can't meet the speed need for the realtime rendering. Improvement on the known software algorithm, using parallel cluster and design dedicated architecture are the three ways to solve the volume rendering accelerate problem. After study the Ray Casting algorithm and its property for parallel computing, we divide the procedure of volume rendering into different tasks module. After that a parallel architecture was described. The main part of dedicated architecture was made up of many image processors. The volume processor in the image processor is responsible for the core algorithm for the volume rendering.Compared with the ordinary hardware accelerate architecture, parallel volume rendering architecture must handle the assignment of volume data among different process unit. The volume data was divided into subcubes, each subcube was marked with an address and assigned to a processor based on some transformation on its address. Inside each subcube, volume data was addressed. These two-layer address assignment scheme made up of the assignment of volume data to different processor to get the eight volume data needed in the trilinear interpolation from memory in a single cycle.Another problem for parallel architecture of volume rendering is the interconnect method among different processors. In the parallel computing system architecture, the building blocks are interconnected by high-speed, scalable and good algorithm adapting networking technologies. The multistage interconnection network is one of the most important technologies of interconnection networks, which has good performance, low price and low complication. Omega is one of the multistage interconnection networks, which has equivalence in topology with the hypercube. Simple Omega Network is a kind of blocking network. This thesis use an expendably non-blocking routing algorithm based on the Omega Network to solve the interconnection problem. EDA technology was used to simulate and verify the design.