Benchmarks
Live BenchmarkDotNet results for every collection in Celerity, measured on each main commit and compared against its .NET BCL counterpart. Hosted CI runners are noisy — treat trends as the signal, individual points as noise.
Lookup
—
vs Dictionary / HashSet
Insert
—
vs Dictionary / HashSet
Remove
—
vs Dictionary / HashSet
.NET BCL baseline
Celerity
Methodology
What you're looking at. Each chart plots one operation (e.g. Lookup) on one collection, with the BCL baseline and Celerity's implementation over time at 100,000 items. Y-axis is milliseconds per invocation, lower is better — click any chart for a full-size view with nanosecond precision. Numbers come from BenchmarkDotNet running in GitHub Actions on
ubuntu-latest with WarmupCount=3, IterationCount=5. The full source is in src/Celerity.Benchmarks. Local Release runs use the default BenchmarkDotNet job and produce more stable numbers — this dashboard exists to catch regressions and surface trends, not as a precision instrument.