Our Research Group Wins the 2016 ACM Gorden Bell Prize
On November 17th, 2016 Supercomputing Conference (SC2016) was held in Salt Lake City of America. The “10M-Core Scalable Fully-Implicit Solver for Nonhydrostatic Atmospheric Dynamics” was awarded the “Gordon Bell Prize”. This achievement was accomplished jointly by several groups leading by Yang Chao from Institute of Software Chinese Academy of Sciences, Xue Wei from College of Computer Sciences Tsinghua University, Fu Haohuan from Center for Earth System Science Tsinghua University, and Wang Lanning from Beijing Normal University. This is the first time that Chinese scientists won this award since its establishment 30 years ago, which is a landmark in the development of China high performance computing application. As the two of the Corresponding authors, professor Haohuan Fu and postdoctoral Lin Gan in our research group plays the key role in this project.
This achievement has been accomplished jointly by Institute of Software Chinese Academy Science, College of Computer Sciences and Center for Earth System Science of Tsinghua University, Climate Change and Earth System Sciences of Beijing Normal University, Chinese National Parallel Computer Engineer Technology Research Center and National Supercomputing Center in Wuxi and other institutes. The achievement was published in Supercomputing Conference this year, titled with “10M-Core Scalable Fully-Implicit Solver for Nonhydrostatic Atmospheric Dynamics”. And the co-corresponding authors are Yang Chao, Xue Wei, Fu Haohuan and Gan Lin, and there are 7 out of 12 authors from Tsinghua University, including Professor Zheng Weimin and Yang Guangwen (the director of National Supercomputing Center in Wuxi) from college of Computer Science.
This research developed a new scalable fully-implicit solver and software applied on atmospheric dynamics simulation. This is the first time in the world that we can accomplish ten million core scalable parallel computing in large-scale heterogeneous many-core system (Sunway Taihu Light), and is also the first time that we can finish an atmosphere simulation with a grid resolution less than 500m in a limited time scale.
The achievement has great breakthrough in application and solver. In application, this results approved that fully-implicit solver is a competitive choice for establishing future super-resolution atmosphere model. The computing efficiency of simulating atmospheric dynamics processes improves about 1 order of magnitude than American next generation atmospheric global climate model (GFDL developed AM3 nonhydrostatic atmospheric dynamics). This cheering results can be applied on high resolution climate model and high-accuracy numerical weather prediction to improve the accuracy of prediction and forecast ability. On the other hand, in solver, this research developed the first implicit solver which can be scaled to ten million cores and whose peak efficiency can be greater than 6%. This degree of parallelism and peak efficiency are 1 order of magnitude higher than those of the work awarded “Gordon Bell Prize” in 2015. This achievement can be fully used in future astronomy, earth science, energy and other fields with the aspect of challenging computing problems.
ACM “Gordon Bell Prize” was set in 1987 by Association for Computing Machinery (ACM). And it will be awarded in the Supercomputing Conference held every November. It is aimed at awarding the latest research results in parallel computing in the world, especially in the field about high performance computing innovation application.
In the meantime, on November 15th, Sunway Taihu Light, operated by Tsinghua University, was ranked first in top 500 list announced in this year’s Supercomputing Conference. On November 16TH, Sunway Taihu Light was awarded the Global Network Leading Technology Achievement prize in the third world conference on Internet in Wuzhen.