ISC High Performance 2023 (ISC 2023) is an international conference and exhibition for a global high-performance computing community of technology providers and users. The final results of the Student Cluster Competition of the ISC 2023 were recently announced in Hamburg, Germany, and the team GeekPie_ from ShanghaiTech University won third place in the online competition. The first and second place teams were from Nanyang Technological University and the University of California Santa Cruz.
There were online and onsite competitions, and fifteen teams from seven countries were offered the opportunity for the online finals. The team GeekPie_ consists of six undergraduates from SIST. They are Li Zecheng ’23, Li Zongze ’24, Hu Aibo ’25, Zhang Yishi ’25, Huang Lei ’25 and Wen Xuanjun ’25. The instructor is Associate Professor Fan Rui.
Group photo of the team GeekPie_. The first from the left is Fan Rui
For the online competition, all the teams were required to run Quantum ESPRESSO, FluTAS, and POT3D applications on two different computing clusters provided by the organizer. Before the competition began, the ShanghaiTech team made full preparations for the competition. They had carefully studied the clusters environment, investigated the application scenarios of the specified software, and prepared the basic software packages. In addition to the preparations, the knowledge in computer architecture and operating system they had accumulated in their courses also laid a solid academic foundation for the final award.
High-performance computing applications often rely on basic libraries such as linear algebra libraries, cross-process communication libraries, and thread parallelism libraries, and there are different commands in each library. The difficulty of the competition is that each team needs to choose the right dependency library to ensure the correctness and efficiency of the program. Furthermore, because the hardware architecture of the two clusters is different, the team needs to fine-tune the software, such as the parallelism and communication strategies, so as to achieve optimal performance on the two clusters. During the competition, the ShanghaiTech team analyzed the performance of each application, found out the hot codes and carried out targeted optimization, and improved the operating efficiency of the software. The final score was determined by the performance of the optimized software and the interview results of the judges. After fierce competition, the team GeekPie_ ranked third, showing the strong strength of ShanghaiTech in supercomputing education and research.