Benchmarks

The benchmarks are built with criterion.rs.

Measurements are taken on the elapsed wall-time.

The benchmarks only focus on sucessfull transactions and vps: in case of failure, the bench function shall panic to avoid timing incomplete execution paths.

In addition, this crate also contains benchmarks for WrapperTx (namada::core::types::transaction::wrapper::WrapperTx) validation and host_env (namada::vm::host_env) exposed functions that define the gas constants of gas (namada::core::ledger::gas).

For more realistic results these benchmarks should be run on all the combination of supported OS/architecture.

Testing & running

To enable tracing logs, run with e.g. RUST_LOG=debug.

To ensure that the benches can run successfully without performing measurement, you can run make test-benches from the workspace run.

To test a selected bench can run successfully on a single run, use can use e.g.:

cargo test --bench native_vps

To benchmark a selected bench with a minimum sample size use e.g.:

cargo bench --bench whitelisted_txs -- --sample-size 10