Evidence / reproducible runs
Benchmark catalog
A transparent view of test health, coverage and performance. Every number carries its run, profile and environment.
PERFORMANCE
Benchmark catalog
micro
select-medium-catalog
35429.0ns/op
74242.0B/op
15.0allocs/op
9.2%CV
go test ./internal/selector -bench BenchmarkSelectMediumCatalog
Limitations
- Compare only with the same profile, dataset and compatible hardware.
micro
record-measurement
17167.0ns/op
224.0B/op
1.0allocs/op
8.5%CV
go test ./internal/metrics -bench BenchmarkRecord
Limitations
- Compare only with the same profile, dataset and compatible hardware.
- BenchmarkRecord is measured once per process because its bounded buffer intentionally changes the cost after saturation; a long calibration run would measure buffer shifting rather than a stable operation.
persistence
flush-persisted-measurement
16188756.0ns/op
6016.0B/op
167.0allocs/op
36.1%CV
go test ./internal/metrics -bench BenchmarkFlushMeasurement
Limitations
- Compare only with the same profile, dataset and compatible hardware.
micro
plan-layers-medium-dag
249781.0ns/op
162773.0B/op
303.0allocs/op
3.9%CV
go test ./pkg/contract -bench BenchmarkPlanLayersMediumDAG
Limitations
- Compare only with the same profile, dataset and compatible hardware.
load
run-plan-concurrent-medium-dag
320015458.0ns/op
3000776.0B/op
9655.0allocs/op
2.6%CV
go test ./internal/orchestrator -bench BenchmarkRunPlanConcurrentMediumDAG
Limitations
- Compare only with the same profile, dataset and compatible hardware.