Open Source Friday with Flash-X: a Multiphysics Simulation Software

()

FlashX Software

FlashX is an open-source instrument used for scientific simulations, funded by the Department of Energy (DOE) and Argonne National Laboratory.
It simulates physical phenomena using different GitHub libraries and can solve complex partial differential equations.
FlashX is used in scientific workflows involving theory, experiments, data generation, analysis, machine learning, and scientific knowledge extraction.
It has a more robust software design compared to other open-source and commercial software, making it easily extensible to different domains.
FlashX uses Fortran, C++, and Python and runs simulations on supercomputers using different nodes and hardware (CPU and GPU).

To access the private FlashX repository, users need to request access through GitHub or by sending an email to anl.gov.
FlashX follows a laboratory notebook policy where developers create notebooks for specific problems and share them with collaborators.

To run simulations on a supercomputer, you can use a tool called JobRunner.
JobRunner simplifies the process of setting up software stacks and submitting jobs to HPC schedulers.
To set up the software stack, use the commands "jobRunner setup software amrx", "jobRunner setup software flashkit", and "jobRunner setup software flashex".
To set up a simulation, use the command "jobRunner setup simulation [simulation name]".
To visualize the results, use ParaView and connect it to the supercomputer where the simulation was run.
To run the simulation interactively, follow these steps:
- Install requirements.
- Set up software using job runner.
- Submit the simulation.
- Visualize the data using ParaView.

FlashX is being used by universities to train neural networks with simulation data.
It is also being used to study metal casting and Olympic diving.

FlashX is exploring cloud computing and containerization.
The team recommends using containers on a high-performance computing system for optimal performance.
Containerization is a complex process, but the team encourages viewers to try it out and engage with the team by filing issues on GitHub.
The team is open to helping people understand and use their workflows.