Benchmarks

Looking for the indicator library's numbers?

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

A screener's cost is dominated by folding every symbol's history through its indicators and evaluating the condition tree at each bar. The benchmarks here measure that core scan work, so throughput scales predictably with the universe size and the number of indicators a spec references.

What is measured

The wickra-screener-bench crate (criterion) covers scan_batch across a matrix of:

• Universe size — 100, 1 000 and 10 000 symbols.
• Indicator count — specs referencing roughly 5 and 20 indicators.
• Execution path — the default rayon parallel feature vs --no-default-features (the sequential / WASM path), which must produce a byte-identical report.

Methodology

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

cargo bench -p wickra-screener-bench                        # parallel (rayon), the default
cargo bench -p wickra-screener-bench --no-default-features  # sequential, the WASM path

The nightly bench.yml workflow runs both on a clean Linux runner and uploads the two together, so the cost of the sequential path is tracked beside the parallel one rather than assumed.

Results

Measured with cargo bench -p wickra-screener-bench (criterion) on a Windows x86-64 laptop, default parallel (rayon) path. The sequential figures are not tabulated here because they depend far more strongly on core count than the parallel ones do; the nightly run reports both, and that artifact is the place to read them. Figures are the median estimate; treat them as orders of magnitude, not guarantees — they vary with CPU core count and toolchain.

Benchmark	Universe × indicators	Median	Throughput
scan_batch/100sym_5ind	100 × ~5	1.34 ms	~75 K sym/s
scan_batch/100sym_20ind	100 × ~20	5.32 ms	~19 K sym/s
scan_batch/1000sym_5ind	1 000 × ~5	12.7 ms	~79 K sym/s
scan_batch/1000sym_20ind	1 000 × ~20	51.3 ms	~20 K sym/s
scan_batch/10000sym_5ind	10 000 × ~5	126 ms	~80 K sym/s
scan_batch/10000sym_20ind	10 000 × ~20	513 ms	~19 K sym/s

The takeaway: per-symbol throughput stays roughly constant as the universe grows (~80 K symbols/s at 5 indicators, ~19 K at 20), so scan cost scales linearly with universe size and with the number of distinct indicators a spec references — a 1 000-symbol, 5-indicator screen finishes in ~13 ms. The nightly bench.yml workflow reruns this on a clean Linux runner for tracking over time.

Caveats

These figures bound the screener's own scan overhead only. End-to-end time in a real run also depends on loading the universe 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.