|
|
| As the semiconductor technology grows with the Moore's Law |
 |
Text Size: A A A |
|
As the semiconductor technology grows with the Moore's Law, how to efficiently utilize the huge amount of transistors available on chip to design high performance processors has become one of the hottest research area. The requirements of the technology and applications to the future high performance CPUs can be summarized with "two HIGHs and two LOWs" which are high efficient, high reliability, low power consumption and low cost.To reach these goals, the following critical scientific problems need to be addressed: 1) what's the relationship between the efficiency of information processing and the complexity of information processing systems? 2) what's the relationship between the efficiency of information processing and the necessary energy to process the information? 3) on top of chips with billions of transistors, how to build systems with reliability and predictable performance?
This project focuses on the above three scientific problems and intends to find out new principles, new architecture and new technology for the microprocessors after the Moore's Law has been extended. Specifically, the following respects will be main concerns: 1) scalable and reconfigurable on chip parallel architecture, 2) the programming model and the supporting environment of the on chip parallel system, 3) the architecture and the circuit design technology for low power design, power management, 4) the reliability design of high performance processors, 5) the design, verification and testing of high performance processors.
The outcome of this project is expected to be an integrated achievement: to implement a teraflops processor prototype with 45nm technology to demonstrate the research results in all the respects.
This project will provide technical background and human resource for the high performance processor design in 12th five-year plan, and will have great impact to the development of IT industry.
|
|
|
