NSFC Key Project: Research on Hyper Parallel Processing Architecture
Grant No. 60633040
PI: Professor SUN Ninghui
The trend of high-productivity microprocessor technology is integrating more processor cores, using more coarse-grained thread-level and data-level parallelism to improve the performance and reduce complexity, power, cost at the same time. When building high performance computer based on multicore CPU chip, the chip level parallelism will be added to the system. Thus will form a new chip-node-system three-level parallel architecture, each level has certain complexity. The total scale will reach more than 10 thousands processor cores. According to this trend, the hyper-parallel processing architecture will be researched in this project. The new parallel computing model, the balance design of computation, storage, communication, interconnection and I/O infrastructure, as well as operating system, programming model, application optimization will be studied. Though theoretical model, full system simulation and application benchmarking the system productivity will be evaluated and verified. The research work of this project will provided theoretical and experimental suggestions and key technologies to the next generation peta-flops high productivity computer architecture.