The winner of the overall championship prize and the e-Prize in the Asia Supercomputer Community's 2016 Student Supercomputer Challenge was the Huazhong University of Science and Technology team.
The researchers earned the e-Prize for optimizing a deep neural network (DNN) program to create an extremely accurate training model for about 600,000 speech data segments in English, Chinese Mandarin, and the Sichuan Dialect, improving computing performance by a factor of 108. The optimization was conducted on eight nodes of the Tienhe-2 supercomputer, with the participation of Intel Xeon Phi co-processors being a unique aspect of the contest.
Some of the student teams have access to Phi hardware at their base institution or a sister organization, while others were able to buy a node to experiment with, and others were using them for the first time at the competition. Students were able to practice the DNN challenge with a sample dataset of 15,000 segments of speech data.
Other apps the competition mandated included the benchmark standard High Performance Conjugate Gradients, a surface wave numerical model MASNUM, and a material modeling software ABINIT. The "mystery app" was ABySS, a de novo, parallel sequence assembler designed for short paired-end reads and large genomes.
From HPC Wire
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found