Benchmarks

Looking for the indicator library's numbers?

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

A radar's cost is dominated by folding a perp universe (open interest, funding, order-book and liquidation events) into per-signal scores and aggregating a RadarAlert across every symbol. The benchmarks here measure that core scan work, so throughput scales predictably with the universe size and the number of signals a spec references.

What is measured

The radar-bench crate (criterion) covers a scan across a matrix of:

• Universe size — the number of symbols and the number of events per symbol folded before the alert is built.
• Signals — specs enabling one signal vs all five.
• Mode — a full batch scan vs the streaming feed + alerts path.

Methodology

Run against fixed, in-process synthetic universes so the numbers are reproducible and contain no I/O variance:

cargo bench -p radar-bench

Results

Measured on one developer machine (release build, parallel feature), median criterion estimates. Treat these as orders of magnitude, not guarantees — they vary with CPU and toolchain.

A full batch scan with the composite spec (all five signals) over a synthetic universe:

Universe	16 events/perp	64 events/perp
100 perps	~0.43 ms	~0.45 ms
1,000 perps	~3.7 ms	~3.8 ms
10,000 perps	~37 ms	~40 ms

The scan is roughly linear in the number of symbols — 10× the universe is about 10× the time — because each symbol folds independently. It is nearly flat across events-per-perp: each signal reads a small fixed window, so most of the per-symbol cost is fixed overhead rather than proportional to the stream length. At 10,000 perps × 64 events (640k events) a full five-signal scan runs in ~40 ms — on the order of 16M events/second.

Caveats

These figures bound the scan overhead only. End-to-end time in a real run also depends on loading the events from disk or a live feed, which these in-process benchmarks do not capture.

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