Abstract of CIPAH Project

Experiment and modelling are essential components of the fusion energy science research program with a key overarching goal being to reach a level of scientific understanding to enable an accurate predictive capability of burning tokamak plasmas. There are many prominent and urgent scientific issues impacting the burning plasma program and the successful operation of the ITER experiment. The parallel simulation codes that address these questions are becoming more and more complex: they are composed of a growing set of sub-models, use more and more complex numerical and parallelization schemes, contain a rather large number of lines of code and accept an increasing number of input parameters. Tests should be devised to detect bugs that might lead to crashes as early as possible so as to ease their correction. Additionally, the performance of these codes has to remain at a high level so as to get results in a reasonable time on the available computers.

In this project, we will consider two simulation codes: JOREK (Edge Localized Modes and disruptions), GYSELA (Ion Temperature Gradient driven turbulence in the core). The goal is to port their Non-Regression Testing suite onto Helios. These tests should be automated and triggered either each time the code repository is modified or periodically (each week-end typically). In this proposal, the work done will permit to increase the confidence level in two parallel codes that are already used in production on Helios. Furthermore, the monitoring, over time, of the runtime of representative runs on Helios, will improve the performance of these codes because developers will receive notice as early as possible of the performance degradation.