Benchmarks

Looking for the indicator library's numbers?

This page is about Wickra Compile. Wickra's own indicator benchmarks — the comparison against TA-Lib, talipp, pandas-ta and the other Rust TA crates — live at wickra.org.

Criterion micro-benchmarks for the codegen hot paths, in crates/compile-bench. Reproduce with:

cargo bench -p compile-bench

The figures below are indicative single-machine numbers (developer laptop, Rust release build); treat them as orders of magnitude, not a spec. The nightly bench.yml workflow re-runs the suite and uploads the report.

Operation	What it measures	Time
spec_hash	Canonical serialisation + SHA-256 of a StrategySpec.	~13 µs
project_hash	Hashing the generated file set in path order.	~10 µs
generate (wasm)	Full codegen: validate → render → manifest, no cargo.	~43 µs
generate (binary)	Same, targeting a native binary.	~42 µs
generate (no_std)	Same, targeting a no_std MCU.	~47 µs
manifest_of	Manifest for a spec without materialising files.	~46 µs

Code generation is cheap — tens of microseconds — because it is pure data templating and hashing with no compilation. The optional cargo build step dominates end-to-end time by orders of magnitude and is deliberately not part of these microbenchmarks; it depends on the target toolchain and is measured separately.

The numbers above are the ones in the repository's BENCHMARKS.md, measured with the commands it names.