This Dataset contains the setup and the results of the convergence tests which are reported in the GALÆXI Paper (Section 5.1). The results are contained in the file results.txt. The used case is based on the method of manufactured solution. The detailed formulation is specified in Hindenlang et al. and is implemented as ExactFunc=4 in GALÆXI.
The folder convtest/ contains the setup for the simulations in the format required by the Reggie2.0 tool (available on GitHub) to run the different cases in an automated fashion. The so-called userblock provided by GALÆXI can be used to rebuild the exact code versions to obtain the results from the paper. To build these versions employ the provided Python build script as:
python build.py ./build-folder ./userblock.txt
Moreover, the script build.sh automatically clones GALÆXI from GitHub, builds the code versions used for the paper and runs the convergence tests using the Reggie2.0 tool. For this, run
bash run.sh
Note: Please ensure that all necessary dependencies of GALÆXI are available (including CUDA) and a Python3 environment is installed on the system. Moreover, the cases can become rather large, so that a consumer GPU might not be able to run the largest cases causing them to be skipped. The largest meshes can be excluded by removing them from the individual parameter.ini files in the convtest/ folder.
This work was funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Sweden, Germany, Spain, Greece, and Denmark under grant agreement No 101093393. Moreover, this research presented was funded by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy EXC 2075 - 390740016, by the DFG Rebound - 420603919, and in the framework of the research unit FOR 2895. We acknowledge the support by the Stuttgart Center for Simulation Science (SimTech). The authors gratefully acknowledge the Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) for funding this project by providing computing time through the John von Neumann Institute for Computing (NIC) on the GCS Supercomputer JUWELS at Jülich Supercomputing Centre (JSC) as well as the support and the computing time on “Hawk” and its “Hawk-AI” extension provided by the Supercomputing Centre Stuttgart (HLRS) through the project “hpcdg”. This work was completed in part at the Helmholtz GPU Hackathon, part of the Open Hackathons program. The authors would like to acknowledge OpenACC-Standard.org, JSC, HZDR, and HIDA for their support.