Benchmark history
Measured cross-run trends are advisory: hosted runner noise and temporal drift limit conclusions. Use a controlled same-run comparison for change attribution. Dry and incomplete runs are excluded.
Loading measured history…
| Release / checkout | UTC date | Mean ns/op | StdDev / samples | Ops/s | Bytes/op | Evidence |
|---|
Selected environment and harness provenance
Enqueue-only allocation is unsupported because deferred execution and iteration setup prevent valid attribution. Missing measurements are shown explicitly. Separate series retain different harness, runtime, CPU, quota, affinity and image configurations. Raw records include warmup and actual measurements, full environment, source tree identities and job settings.