Cross-platform authorization benchmarks for the Cedarling bindings,
regenerated on pushes to main that change jans-cedarling/**,
or on a manual workflow_dispatch run against main.
Last updated: 2026-11-19 06:09 UTC (commit f976436).
Each binding is measured on its own GitHub-hosted ubuntu-latest runner.
Runners are provisioned fresh per binding, so cross-binding comparisons are
unpaired and subject to per-runner variance; read cross-binding ratios as
directional rather than exact, especially for closely-matched bindings.
Warm-up iterations per scenario: 100; measured (sample count) per scenario: 1000.
Note:unsigned_batch_* and multi_issuer_batch_* measure the full batch call, not per-item latency.
Relative speed per scenario (x vs Rust, lower = faster)#
Ratios use each binding's full-precision mean_ns over Rust's, so they may not match dividing the rounded µs values shown above. Rows where Rust is unavailable use the fastest binding as the baseline.
Scenario
c
go
python
rust
wasm
multi_issuer_2_tokens
1.47x ███████
1.18x █████
0.89x ████
1.00x █████
2.16x ██████████
multi_issuer_3_tokens
1.39x ███████
1.11x ██████
0.87x █████
1.00x █████
1.88x ██████████
multi_issuer_batch_10
1.39x ████████
0.99x █████
0.84x █████
1.00x ██████
1.81x ██████████
multi_issuer_batch_25
1.41x ███████
1.01x █████
0.84x ████
1.00x █████
1.88x ██████████
multi_issuer_sig_status
skipped
skipped
skipped
skipped
skipped
unsigned_batch_10
1.70x ███████
1.51x ██████
0.91x ████
1.00x ████
2.50x ██████████
unsigned_batch_25
1.54x ███████
1.43x ███████
0.88x ████
1.00x █████
2.11x ██████████
unsigned_deep_json
1.49x ██████
1.20x █████
0.84x ████
1.00x ████
2.32x ██████████
unsigned_simple
1.81x ███
1.71x ███
0.90x ██
1.00x ██
5.99x ██████████
Latency columns (Mean, p50, p95, p99, Min, Max) in the detail tables below are microseconds (µs).