Tsinghua University's supercomputing team won the SC 2018 championship and swept the championships of the three major international college supercomputing competitions

Leifeng.com AI Technology Review: On November 15, local time in the United States, the 2018 International Student Supercomputing Competition (SC18) came to an end in Dallas. The supercomputing team of the Department of Computer Science of Tsinghua University won the championship with a total score of 88.398 points (out of 100 points), 11.518 points higher than the second place Nanyang Technological University of Singapore. This means that the supercomputing team of Tsinghua University successfully won the championship of the three major international student supercomputing competitions in 2018, ASC, ISC and SC, achieving another "Grand Slam" after 2015. The supercomputing team of Tsinghua University has won a total of 11 championships in these three major international student supercomputing competitions.

The Supercomputing Conference (SC) is a top international conference in the field of supercomputing with a huge international influence. As an important part of the conference, the SC competition is a top event in the field of supercomputing. It is held annually and attracts many universities from countries and regions around the world to participate. The competition aims to promote the connection between college students and the industry, promote the exchange and training of young supercomputing talents from all regions of the world, and enhance the level of supercomputing application and R&D capabilities.

A total of 15 universities around the world formed teams to participate in this competition, and Tsinghua University was the only team from a mainland university. It faced fierce competition from traditional powerhouses such as Nanyang Technological University in Singapore and National Tsing Hua University in Hsinchu, Taiwan.

The competition requires participating teams to build their own computing clusters under the restriction of 3000 watts of total power and conduct performance competitions of 6 applications on the cluster system. In the 48-hour competition, the basic performance tests of the supercomputing cluster, HPL (literally translated as High Performance Linear System Software Package) and HPCG (literally translated as High Performance Conjugate Gradient), large-scale machine learning, steady-state solution and simulation of nuclear fission chain reaction, paper reproduction - super-large earthquake simulation, mysterious applications announced on site, etc. At the same time, the competition also set up interviews, conferences, poster design and other links. Members need to introduce their optimized applications and ongoing software optimization designs to the judges during the competition, and they also need to complete an English paper that can be published in an international authoritative magazine within 48 hours.

Based on the pre-match training and the basic performance test of the on-site supercomputing cluster, the Tsinghua team decided to use a cluster server configuration that combines different nodes of GPU (graphics processing unit) and CPU (central processing unit) according to the given load characteristics of some applications, which well balanced the final score of the performance test and the application. In the end, with a comprehensive and profound understanding of supercomputing systems and applications and excellent performance optimization capabilities, the Tsinghua team achieved excellent results in 6 applications and scored 88.398 points, leading the second place Nanyang Technological University of Singapore by 11.518 points. Leading by a large score, it became the overall champion of this year's SC18. Another award of the competition, the highest LINPACK (literally translated as high-performance linear system software package) award, was won by Nanyang Technological University of Singapore.

Medals

The Tsinghua student supercomputing team participating in the competition is mainly composed of six undergraduate students from different grades, including Yu Jiping (52nd year in Computer Science), Yu Xinjian (55th year in Computer Science), He Jiaao (62nd year in Computer Science), Zheng Liyan (64th year in Computer Science), Zhao Chenggang (75th year in Computer Science) and Lou Chenyao (60th year in Computer Science, Yao Class) from the School of Interdisciplinary Information Sciences. They are technically supported by postgraduate students Cao Huanqi, Feng Guanyu and Wang Miao from the Institute of High Performance Computing who have participated in many international competitions. The instructors are Associate Professor Zhai Jidong and Postdoctoral Fellow Han Wentao from the Department of Computer Science. At the same time, five undergraduate students Tang Shizhi (53rd year in Computer Science), Chen Yu (54th year in Computer Science), Gu Yuxian (75th year in Computer Science), Chen Shengqi (63rd year in Computer Science) and Zhang Chen (71st year in Computer Science) also participated in the training.

Team photo

The Tsinghua University Computer Science Department Supercomputing Team was founded in 2012. So far, more than 50 students have participated in 17 international competitions and won 11 championships. The team's 11 championships are not only due to the team members' deep understanding of the application of related scientific fields, solid basic skills in high-performance computing, excellent on-site response strategies and language expression skills, but also the support of Tsinghua University Computer Science Department, which has a complete teaching system in high-performance computing, parallel systems and computer systems, and the guidance of instructors such as Associate Professor Zhai Jidong, who has achieved many scientific research results in the field of supercomputing.

Source: Tsinghua University https://mp.weixin.qq.com/s/iW_AEAE--pn1iuYAjltvNA

What can defeat Tencent is not its stock price, but itself.

Samsung AI processor is here to compete with Apple and Huawei

Luo Yonghao was angry about "Hammer Life and Death"