# Performance analysis of Volna-OP2 -- massively parallel code for tsunami modelling

**Authors:** Daniel Giles, Eugene Kashdan, Dimitra M. Salmanidou, Serge Guillas, Frédéric Dias  
**arXiv:** [2002.04889](https://arxiv.org/abs/2002.04889)  
**Date:** 2020-02-12  
**Summary page:** https://www.summarizepaper.com/en/paper/2002.04889v1/

## Key points

- Volna-OP2 is a software package for simulating tsunamis using HPC architectures
- Recent study focuses on error analysis and scalability of the GPU version of the code
- Authors explore a novel decomposition technique for numerical errors in tsunami codes
- Errors are broken down into dispersion and dissipation components, providing insights into amplitude smearing and phase lagging/leading phenomena
- This approach represents the first time such errors have been assessed in this manner
- Volna-OP2 is widely adopted due to its computational efficiency and ability to leverage HPC architectures
- Study offers guidance on achieving an acceptable level of accuracy within specific time constraints
- Provides valuable insights into the performance of Volna-OP2 as a massively parallel code for tsunami modeling

## Summary

The software package Volna-OP2 is a powerful and efficient code designed for simulating the complete life cycle of a tsunami using state-of-the-art High Performance Computing (HPC) architectures. In a recent study, authors Daniel Giles, Eugene Kashdan, Dimitra M. Salmanidou, Serge Guillas, and Frédéric Dias present a comprehensive analysis of the GPU version of the code, focusing on error analysis and scalability. One key aspect of this study is the exploration of a novel decomposition technique for numerical errors in tsunami codes. The authors break down these errors into dispersion and dissipation components which provide valuable insights into amplitude smearing and phase lagging/leading phenomena commonly observed in most tsunami simulations. This decomposition approach represents a new perspective and tool for understanding these occurrences marking the first time such errors have been assessed in this manner. Volna-OP2 has already gained widespread adoption within the tsunami modeling community due to its computational efficiency. Its ability to leverage HPC architectures has enabled researchers to perform various sensitivity analyses and uncertainty quantification studies. However when conducting statistical studies that require numerous simulations there is always a trade-off between accuracy and runtime. To address this challenge the analysis presented in this paper offers guidance to users on achieving an acceptable level of accuracy within specific time constraints. Overall this study provides valuable insights into the performance of Volna-OP2 as a massively parallel code for tsunami modeling. The error analysis and scalability study contribute to our understanding of numerical errors in tsunami simulations while offering practical recommendations for achieving accurate results efficiently.