From January 10 to April 13, 2024, five students from the School of Computer and Information Engineering participated in the ASC World University Supercomputer Competition in Shanghai and won the second prize.

Introduction

The ASC Student Supercomputer Challenge is an international event initiated by the Asian supercomputer community and supported by experts and organizations in Asia, Europe and the Americas. The main objectives of the Challenge are to promote the exchange and training of young supercomputing talents around the world, to improve supercomputing applications and R&D capabilities, to advance supercomputing, and to promote technological and industrial innovation. The first ASC Student Supercomputing Challenge was held in 2012 and has attracted more than 10,000 undergraduate students from around the world.

ASC24

The ASC24 Student Supercomputing Challenge selected 25 teams from more than 300 university teams around the world to advance to the final live finals. The final was held from April 9-13, 2024 at Shanghai University. Competitors in the final included teams from some of the top universities in China, such as Tsinghua University, Peking University, and Shanghai Jiaotong University, making this year's Challenge extremely competitive and difficult. Each team had to design and build their own computational platform on-site and ensure that they could run their assigned application tasks within the 3 kilowatt power consumption limit. In addition, teams were required to present in front of a panel of judges competing for awards including the overall winner, the Silver Award, the top Linkpack Award and other prizes.

The rules of the tournament were also extremely strict, including limitations on optimization methods, stringent requirements for algorithm changes, and strict monitoring of platform power consumption, all of which greatly increased the difficulty teams had in designing and executing their solutions. In addition, there are several specific performance optimization tasks, such as HPL, HPCG and optimization for the open source software OpenCAEPoro, as well as the newly introduced GoMars and LLM inference tasks, which require teams to maximize the computational performance while ensuring the correctness of the results.

Behind the Honors

Instructor Shen Wenfeng provided great support to our team in this ASC24 Student Supercomputing Challenge. Mr. Shen not only gave guidance on the team's strategy and technical layout, but also provided the team with laboratory space and high-performance computer resources to ensure smooth training and testing. With his help, team members Tao Chenlong, Cai Yuyang, Liang Yueyang, Wu Shengwei, and Chen Rui made full use of the resources to drill and practice day after day in the lab. In preparation, the five team members optimized parallel algorithms, 3k-watt supercomputing cluster design and HPL and HPCG program fine-tuning, as well as tested various supercomputing configurations to ensure the best performance in the competition. Their main efforts were placed on ensuring that the various application tasks ran smoothly under strict rule constraints while meeting the 3 kW power consumption limit.