astro-benchmarks

Benchmarking Astrophysical Simulation Codes

Copyright (c) 2025 The AstroBenchmarks Team

How it works

1. Fork the repo and pick a benchmark from benchmarks/.
2. Set up and run the benchmark using your simulation code.
3. Save your benchmark results to a JSON file (per problem specs).
4. Submit a Pull Request with your results in results/<code>/<machine>/<benchmark>/<commit>/.

Each benchmark includes detailed specifications, reference solutions, and evaluation metrics.

Benchmarks test a variety of aspects of the code, including runtime, throughput, memory usage, and accuracy.

Results are compiled and published on the Leaderboard.

Benchmarks

An example benchmark problem is:

XXX

Citing

If you used AstroBenchmarks for research, please cite it as: