Benchmarking

Record the full context

• reachq version or commit, Python version, operating system, and CPU.
• Graph source and type, n, m, density, vertex ordering, and random seed.
• Algorithm, epsilon, omega, refinement configuration, workers, and hop limit.
• Construction time, query time, peak memory, output size, beta, realised bound, hopbound, and approximation error.

Keep claim types separate

Theoretical bounds, microbenchmarks, and end-to-end measurements answer different questions. Do not use beta alone as a practical-performance proxy, and do not report an application benefit without an explicit experiment.

Reproducible scripts live in the repository scripts ↗.