CAREER: An Effective Integration of Research and Education on High-Speed and Energy-Efficient Interconnects for Multi-Core and Multi-Thread Systems

点击翻译按钮获取中文摘要

Proposal ID: 0846053 Title: CAREER: An Effective Integration of Research and Education on High-Speed and Energy-Efficient Interconnects for Multi-Core and Multi-Thread SystemsPI name: Jiangjiang Liu Abstract: The goal of this CAREER project is to improve the speed and energy efficiency of computers that use multi-core systems. There is enormous demand for these multi-core systems in high-performance and data-intensive applications. When information is sent from one core to another, the interconnect devices create a bottleneck. Compressing the information to reduce its size speeds up the transfer over the interconnects and reduces energy consumption. This project will conduct the first comprehensive compressibility analysis of information on all important interconnect components for multi-core systems in the context of real-world application programs. This will help guide the proposed design of compression techniques for interconnects and provide a deeper understanding of how compression should be applied to help interconnect performance and alleviate energy bottlenecks in multi-core systems. Three novel information compression techniques are proposed. The proposed methods can be adapted and applied to address high-speed I/O channel performance and energy bottlenecks.The CAREER project effectively integrates research and education; promotes undergraduate and graduate research; stimulates and equips aspiring computer science undergraduates who are underrepresented or economically challenged in advanced study; triggers initial interest of K-12 students in computing with strong outreach efforts; enhances Lamar?s infrastructure by providing high-performance computing equipment; enhances scientific and technological understanding by disseminating the results of the program to a broad audience through international publications and a dedicated website; and advances and accelerates expansion of high performance computing applications as well as commercial business applications by improving the interconnect speed and overall performance of multi-core systems.