On-going Projects
Universal Multi-Processor (UMP) Structure
We have witnessed that Petascale computing barrier was easily broken by traditional MPP or commodity cluster systems, which is extensively used in Terascale era. However, there is a consensus that we need more revolutions to welcome Post-Petascale and Exascale era since the issues of efficiency, power and reliability are challenging our current techniques. We are devising a novel architecture--Universal Multi-Processor (UMP)--to serve next and future generation supercomputing systems. In our vision, such a supercomputer comprises mulit/many-core components (Godson-T, GPU), billions of threads and non-cache coherent global address spaces. Several issues are addressed:Intelligent Memory System
Coming soon... more...Scalable Algorithms
We have already entered into Peta-scale computing era. It is believed that an Exa-scale system will appear in the next decade. During this evolution, it is important to study co-design of architectures and applications in order to create synergy in their respective evolutions for closing the gap between the peak capabilities of the hardware and the performance realized by high performance computing applications (application-architecture performance gap). In such large-scale supercomputing systems, scalability is involved in not only reaching peak performance, but achieving high reliability. This project focuses on the development of architecture-aware algorithms and domain-specific optimization systems to irregular computations including general sparse linear systems, molecular dynamics and other non-numeric algorithms such as graph, multiple pattern matching and dynamic programming.more...
Simulation Toolkit (SIMK)
New computer system desings involve modification of hardware organization, programming paradigm and algorithm. In the early stage of exploring a computing system, premature of hardware or software often limits the flexibility of the other. An alternative solution is to allow algorithm, programming and the underlying hardware evolve seperately, while allowing them feadback each other. In order to achieve this goal, we have developed a simulation toolkit--SIMK--to explore innovation at all layers of computer system, i.e. microarchitecture, memory, communication. Using SIMK we sucessfully built an experimental platform for verifying the idea of HPP architecture. Now, we are constructing a many-core (Godson-T) supercomputer simulator.more...
Hyper Parallel Processing (HPP) Operating System
Coming soon... more...Network Flow Analysis
Network Flow Analysis (NFA) is designed a real time processing system, which filters the rapidly increasing massive data produced by Internet backbone. With respect to the evolvement of multi-threaded multi-core processors, applications of NFA is adaptive to exploit massive concurrency to improve throughput and reduce response time. We are targeting at a highly concurrency NFA system which take full advantage of multi-threaded multi-core parallel computers.more...
