The rapid development of supercomputers and the growing needs of extreme-scale applications pose grant challenges in the development of highly efficient and scalable parallel algorithms. An indisputable fact is that there is a large gap between the expectations from an algorithm developer and the design space considered by a hardware architect. For example, many-core based heterogeneous architectures have become a competitive choice to build supercomputing systems that run at desired computational rates and fit within an affordable power envelope. These systems can deliver extremely high computing throughput but often come with relatively low capability of data movement. Therefore, it is of crucial importance to adapt the parallel algorithm with the architectural features so that the potential of the supercomputers can be fully unleashed at the extreme scale.
In this talk, the speaker will present his recent work that won the ACM Gordon Bell Prize of 2016, jointly conducted with Wei Xue, Haohuan Fu, et al. (Proc. of SC’16, article 6). In the award-winning work, the speaker and his collaborators have designed a fully implicit solver for solving the atmospheric dynamics at the nonhydrostatic scale with high scalability on state-of-the-art many-core based heterogeneous systems. In particular, they propose a highly efficient hybrid domain-decomposed multigrid preconditioner with a geometry-based pipelined incomplete LU factorization for solving the overlapped subdomain problems. They perform systematic optimizations on different hardware levels to achieve best utilization of the heterogeneous computing units and substantial reduction of data movement cost. The fully-implicit solver successfully scales to the entire system of the Sunway TaihuLight supercomputer, which is the fastest supercomputer in the world, with over ten millions. This is, to the speaker’s knowledge, the largest fully-implicit simulation to date.
About the speaker
Prof Chao Yang received his PhD in Computer Science from the Institute of Software of the Chinese Academy of Sciences (CAS) in 2007. He then joined the faculty of the Institute, where he has been a Professor since 2013. He is also the Vice Director of the Laboratory of Parallel Software and Computational Sciences of the Institute, and a Professor at the State Key Laboratory of Computer Science of CAS. From 2007 to 2014, he was a Research Associate at the University of Colorado Boulder.
Prof Yang’s research interests include numerical analysis and modeling, large-scale scientific computing and parallel numerical software. He has published over 50 papers in peer-reviewed journals and is a reviewer for Mathematical Reviews of the American Mathematical Society (AMS), Journal on Scientific Computing of the Society for Industrial and Applied Mathematics (SIAM), Journal of Computational Physics, and Transactions on Parallel and Distributed Systems of Institute of Electrical and Electronics Engineers (IEEE). He is also a member of IEEE, SIAM and the Association for Computing Machinery (ACM).
Prof Yang is the leading recipient of the 2016 ACM Gordon Bell Prize, the highest honor in high performance computing, awarded at the annual International Supercomputing Conference in Salt Lake City in the United States.
For attendees’ attention
The lecture is free and open to all. Seating is on a first come, first served basis.
