Benchmarking search settings

Benchmarking search settings

Benchmarking is a two-step workflow:

1. Run csubst benchmark to launch multiple csubst search runs on the same input while varying selected settings.
2. Run csubst benchmark-plot to collect existing benchmark summaries and write comparison tables and an overview figure.

This page is a workflow index. The command-specific pages above keep the examples, options, and output-file descriptions in one place.

Suggested workflow

1. Validate the dataset with csubst doctor.
2. Run a small csubst benchmark grid first.
3. Inspect csubst_benchmark/csubst_benchmark_summary.tsv.
4. Add more settings or thread counts only after the small grid behaves as expected.
5. Use csubst benchmark-plot to compare one or more benchmark directories.

Related pages

• csubst benchmark
• csubst benchmark-plot
• Parallel processing and performance tuning