Lin Gan from Tsinghua University
In this video from SC19, Lin Gan from Tsinghua University presents: HPC Solutions for Geoscience Application on the Sunway Supercomputer.
In recent years, many complex and challenging numerical problems, in areas such as climate modeling and earthquake simulation, have been efficiently resolved on Sunway TaihuLight, and have been successfully scaled to over 10 million cores with inspiringly good performance. To carefully deal with different issues such as computing efficiency, data locality, and data movement, novel optimizing techniques from different levels are proposed, including some specific ones that fit well with the unique architectural futures of the system and significantly improve the performance. While the architectures for current- and next-generation supercomputers have already diverted, it is important to see the pro and cons of whatever we already have, and to make up the bottlenecks as well as maximize the advantages. This talk contains the summary of the most essential HPC solutions that greatly contribute to the performance boost in our efforts on porting geoscience applications, the discussions and rethinking, and the potential improvements we could undertake.
Lin Gan is a postdoctoral research fellow in the Department of Computer Science and Technology at Tsinghua University and the assistant director of the National Supercomputing Center in Wuxi. His research interests include high-performance computing solutions to geoscience applications based on hybrid platforms such as CPUs, FPGAs, and GPUs. Gan received a PhD in computer science from Tsinghua University. He has received the 2016 ACM Gordon Bell Prize, Tsinghua-Inspur Computational Geosciences Youth Talent Award, and the FPL Significant Paper award. He is a member of IEEE.
