Performance Analysis of Scinetific Applications on Multiple RISC-V Platforms

This thesis presents a comprehensive performance analysis of a scientific applica tion, SeisSol, executed on multiple risc-v platforms available in the hca cluster at the Barcelona SupercomputingCenter(bsc). Theevaluation is conducted using the methodology developed by the Performance Optimization and Productivity (pop) CenterofExcellence andperformanceanalysistoolsdevelopedatbsc. Theselected application is representative of high-performance computing (hpc) workloads and has been optimized for parallel execution. The performance analysis includes both scalability studies and tracing-based investigations in single-node and multi-node configurations. Detailed execution traces are captured using the Extrae tool, and the resulting traces are analyzed with Paraver and Dimemas to extract relevant performance metrics and identify potential sources of inefficiency. A comparative analysis is also performed to better understand the performance of the application on risc-v architectures compared to traditional x86_64 architectures.