Coordination Of Multiple Processors And Multiple Graphics Hardware 3d Data Field Is Parallel Volume Rendering Research

The direct volume rendering techniques can provide vivid, realistic views of data for analysis purpose in 3D scientific computational field. But the increasing size of volumetric data and image make it impossible to achieve the interactive rendering because of the incoming massive computation. To overcome those problems, graphics hardware assisted and parallel techniques are often been utilized for accelerating volume rendering, but both of them have their shortcoming. Hardware-accelerated volume rendering on single CPU and GPU was mainly restricted to the amount of texture memory required to store the volume data, therefore hardly handle large-scale volume data. Software-based parallel volume rendering will consume too many CPU resources for interactive capability. However, when mixing the two accelerating techniques mentioned above, the parallel volume rendering can then break the bottleneck in memory and performance using multi-processor and multi-graphics hardware.But in multi-processor system, the factor price made the number of graphics hardware hardly coincide with the number of processors. The challenge in this context is to balance multi-processor and multi-graphics hardware for efficient volume rendering, it strongly rely on the rendering algorithm design. So this paper studied the balancing factors in parallel hardware-accelerated volume rendering, the aim of this work includes how to configure the CPU and GPU numbers depending on the visualization requirements, design effective algorithms to minimize the consuming of hardware resources, and well perform volume rendering utilizing remote graphics hardware resources with general parallel machine without graphics module, all of these problem demand to be researched.This work implements the algorithms to balance multi-processor and multi-graphics hardware in parallel volume rendering for 3D large-scale dataset on SGI shared memory multiprocessor system. Experimental results show that our parallel hardware-accelerated volume-rendering algorithm can adapt to local and remote hardware rendering environment, and achieve good interactive rendering frame rates. The contribution of this work include the following aspects:1. A parallel volume-rendering model is presented, which can be used to balance multi-processor and multi-graphics hardware for rendering and the configuration of hardware number on SGI multiprocessor system. During the implement phase, we adopted several key strategies for efficiently balancing the workload exist in processor and graphics hardware, they include: the image quality/rendering-speed trade-off, the data access control from graphics hardware memory to main memory, minimizing message passing during the parallel image compositing and the parallel pipeline technique.2. An architecture of hybrid parallel rendering based on Multiprocessors and distributed graphics workstations is presented. The general parallel machine without graphics module can then perform effective local 3D data volume rendering utilizing remote graphics hardware resources.3. Based on large numbers of Experimental test and the analysis between different data size, we have verified the validity, efficiency and scalability for our balanced parallel rendering model. All the experiment have been tested in the two kind of parallel rendering environment we mentioned in this paper.