Ding Zhiyu

Bachelor

In the months leading up to the conference, teams compete remotely using provided hardware in an education-focused experience supported by HPC experts. A 48-hour contest held concurrently with SC is the culmination of experience and knowledge gained by the teams in the preceding months.

• Optimized MLPerf benchmark on competition machines, boosting efficiency and throughput in machine learning tasks

• Conducted NAMD simulations to study water properties, protein folding, and diffusion coefficients under various conditions.

• Optimized NAMD configurations for large-scale simulations, achieving high simulation efficiency

• MLPerf Optimization
• NAMD Simulations
• HPC Competitions

The final competition uses part of the source code of the self-developed oil spill prediction model as the competition problem. It primarily assesses participants' understanding of the Euler method for solving trajectory equations and their ability to use the vector method to determine whether "oil particles" will be adsorbed onto the shore.

• By using a hybrid MPI and OpenMP parallelization on the original serial program, and utilizing load balancing, the algorithm’s execution efficiency is improved, fully leveraging the 2-node 128-core computing resources.

• Optimization techniques like fast exclusion and binary search quickly determine the relationship between the oil particle trajectory and the coastline.

• Data access order is rearranged using Fortran/C, improving cache hit rates by exploiting memory locality.

Awarded to Zhiyu Ding, PanyiLiu, Zixiang Zhou, Guangyong He from Southwest Petroleum University, under the guidance of Professor Bo Peng and Quan Zhang.

Awarded third prize in the National Finals of Python programming design competition, Blue Bridge Cup, national final for professional software and information technology talents.