Automatic Dynamic Task Distribution between CPU and GPU for VR Systems

摘要：

The increase of computational power of programmable GPU (Graphics Processing Unit) brings more opportunities for using these devices for general purpose computing. Hence, how to use it as a coprocessor properly becomes very important. This paper presents a new architecture to implement automatic and dynamic task distribution between CPU and GPU for VR systems. This architecture is based on FSM (Finite State Machine) which is defined in Lua scripts and used for task distribution. Each state of the FSM containing a set of heuristics represents a state of the running system. The architecture can change the current state to get acquainted with the best processor for handling a given task automatically in complicated and changeable real-time systems. This paper also presents a method to implement particle systems using GPU. Then tests and results of implementing the architecture mentioned above are based on particle systems.

关键词： automatic distribution GPGPU FSM particle system

作者: Shuxin Qin Xiaobing Geng Yongshi Jiang

作者单位: Integrated Information System Research CenterInstitute of Automation, CASHaidian District, Beijing, Integrated Information System Research Center Institute of Automation, CAS Haidian District, Beijing

会议类型: 国际会议

会议名称: 2011 International Conference on Information System and Computational Intelligence(2011 IEEE信息系统与计算智能国际会议 ICISCI 2011)

会议地点: 哈尔滨

会议语种:英文

页码: 1-4

在线出版日期: 2011-01-18（万方平台首次上网日期，不代表论文的发表时间）