Research On GPU High-performance Computing Based On ParaViewWeb Framework

With the development of the mobile terminals such as smart phones and tabletcomputers, many new solutions of traditional problems have been found. Recently,the limited computational capabilities of the mobile terminals become a big barrierthat restraints the migration of applications from PCs to mobile terminals. Therefore,it’s meaningful to propose a new method to improve the performance of the mobileterminals and enable them to do complex computational work. Quantities of complexalgorithms could be applied on the mobile terminals by taking the advantage of themobile internet and the Client/Server model, which can hand over most of thecomputation part to a remote server.Based on the expandability of ParaViewWeb framework, this thesis presents anetwork-based general computing platform by adding capabilities ofhigh-performance general-purpose computing with GPUs while keeping the originalfunctions of3-D data analysis and visualization. This platform provides computationas a service to enable the mobile terminals such as smart phones and tablet computersto do complex computational work. So the GPU computing resources of the servercan be called remotely on this platform. We tested the platform with a medical imagefiltering application.The experimental result shows the significant acceleration of the web-basedapplication compared to the locally running application. With the increasing of thesize of the experimental image, the speed-up of the filtering algorithm of medicalimage ranges from152x to1481x. GPU high-performance computing implementationbased on ParaViewWeb framework reduces the algorithms’ execution time, solves theproblem that mobile terminals were unable to do large-scale operation, and provides abroader future to the mobile devices.