Parallel Distributed Processing And Application In The Rendering Base

Along with the distributed structure and parallel computing technology maturegradually, which use it deal with huge amount of data have become the focus in the industry.Render due to its huge amount of data and longer render time consumption had becomeapplication key in distributed structure and parallel computing technology. The clouds andthe terminal distribution render mode had become application trend, however, with thecomplexity of these two technologies and application scope of growth, building withreal-time interactive ability, high resource utilization rate and low delay distributed parallelrendering platform are facing many technical challenges.This article focus on the research of distributed architecture, parallel computingtechnology and the application of GPU on the following:1. Research of the current situation and demand of rendering platform, analyzes theoriginal render mode of the hardware structure, operation mode and principle, on the basisof distributed parallel rendering platform for further research. Mainly analysis platformoverall framework, including platform system structure, communication model and platformbased on the characteristics of distributed parallel rendering platform for further research.2. Study the relevant technology of the platform, including the calculation of platformstructure type, parallel computing type, visit deposit type and on the basis of the extendedout of CUDA parallel architecture. For CUDA research its outstanding ability of calculationprinciple, programming model, data and instruction loading and storage model.3. Study the basic knowledge of rendering, including the basic process, scene said wayand basic3d transformation model, on the basis of the JPEG format around to terminalrender theory research.4. High delay in render platform were studied, Firstly analyzes the network architectureof existing rendering platform and high delay reason, on the basis of the extended Infinibandsolutions; Then research Infiniband system architecture and address resolution principle;The further research is about Infiniband I/O architecture and traditional server I/Oarchitecture comparison and analyzes the main reasons of the Infiniband advantage thenresearch Infiniband system architecture and address resolution principle, on the basis offurther analyzes the RDMA technology and kernel bypass model.5. According to the current situation on the rendering algorithm was improved,research on the rendering universal model and analyzes render in the situation of shortage, on the basis of trying to build a GPU rendering model, improving rendering model throughthe using of parallel computing way to give full play to the GPU architecture of memoryfunction, finally to save bandwidth and reduce the additional memory flow, and finallyachieve the effect of faster.6. The detailed design of distributed parallel rendering platform realization process,including requirements analysis, design objectives, platform structure, resource scheduling,task decomposition and coding, etc. Given the overall design process and terminal platformsecond rendering algorithm design thought, at last, through the experimental analysis whenactually interactive ability, resources utilization and ductility, the validation based onparallel distributed processing for the improvement of the original render mode.