Automated Tuning of HPL Benchmark Parameters for Supercomputers

This research presents an automated tuning approach for optimizing parameters of the High-Performance Linpack (HPL) benchmark, which is crucial for assessing supercomputer performance. Leveraging the Periscope Tuning Framework (PTF), the study explores the impact of various parameters such as problem size, block size, and process grid configuration on computational efficiency. Through rigorous experimentation and analysis, the proposed plugin efficiently identifies optimal configurations, offering valuable insights for enhancing supercomputer performance.