ConsoleReporter

ConsoleReporter renders results to the terminal as a colour-coded table using Spectre.Console. It is part of the NBenchmark.Reporters.Console package.

Setup

dotnet add package NBenchmark.Reporters.Console

using NBenchmark.Reporters.Console;

.WithReporter(new ConsoleReporter())

CLI usage

When the NBenchmark.Reporters.Console package is referenced, ConsoleReporter self-registers via [ModuleInitializer] and becomes available through the --reporter console CLI flag:

dotnet run -- --reporter console

No explicit .WithReporter(new ConsoleReporter()) call is needed when using the CLI - the host discovers it automatically through ReporterRegistry.

Example output

── BENCHMARK RESULTS  2026-06-06 03:40:00 UTC ──────────────────────────────────

  Benchmark              Median   Mean     Ops/s       Ratio                  Sig    Mag     Alloc/op
  Compute                300 ns   275 ns   3.64 Mops/s ████████ 0.75x         ✓      lrg     -
  Baseline (baseline)    400 ns   376 ns   2.66 Mops/s ████████████ baseline  -      -       -

── Precision & Tail Latency ────────────────────────────────────────────────────
... (error/stddev/cv/dynamic percentile columns)

── Interpretation ──────────────────────────────────────────────────────────────
Omnibus: not run (fewer than 3 comparable groups)
Significance: Mann-Whitney U (p < 0.05)
Outliers: IQR fence (1.5×)
Effect metric: Cliff's δ (Romano neg/small/med/large labels)
Profile: realistic (no per-iteration GC, no between-benchmark GC, alloc tracking on)
2 benchmark(s) · 0.0s total · CI 95%

Compute: auto-tuned: 190 samples × 1 ops, warmup 40, CI ±1.8%
Baseline: auto-tuned: 190 samples × 1 ops, warmup 40, CI ±1.9%

── Warnings ────────────────────────────────────────────────────────────────────
... (only shown when present)

When there are two or more benchmarks, a bar chart of median timings is also displayed below the table.

When three or more benchmarks are compared, the per-row Sig column shows the post-hoc pairwise verdict (candidate versus baseline, Holm-Bonferroni corrected) and a single omnibus line is printed above the footer, summarising the Kruskal-Wallis verdict across all groups:

Omnibus Kruskal-Wallis across 3 groups: H(2) = 7.20, p = 0.027 → significant

Significance: Kruskal-Wallis (p < 0.05)
Outliers: MAD (3×)
Effect metric: Cliff's δ (Romano neg/small/med/large labels)
Profile: realistic (no per-iteration GC, no between-benchmark GC, alloc tracking on)
3 benchmark(s) · 0.0s total · CI 95%

After the Interpretation section, ConsoleReporter prints a grey auto-tuned: … line per benchmark summarising what the adaptive measurement loop resolved - the measured-sample count, ops-per-sample (K), warmup length, and the achieved CI half-width. Pinned runs still show the line, with the resolved counts you set.

After the comparison and precision tables, ConsoleReporter prints an Interpretation section with omnibus/significance context, outlier mode, effect-metric semantics, and the measurement profile. If warnings exist, they are shown in a separate Warnings section below the auto-tune lines. The final summary line shows benchmark count, total run time, and confidence interval.

Columns

Column	Description
Benchmark	Name of the benchmark. Colour-coded: green (≤ 5% slower than baseline), yellow (≤ 50% slower), red (> 50% slower). Baseline is shown in bold.
Median	Median timing.
Mean	Arithmetic mean.
Ops/s	Mean operations per second (1e9 / Mean when timing is in nanoseconds). - for errored or dry-run results.
Ratio	Visual bar plus ratio relative to the baseline. Green for faster, yellow for moderately slower, red for significantly slower. The baseline cell shows baseline.
Sig	✓ = difference from baseline is statistically significant; ✗ = not significant; - = not applicable (baseline or significance not tested).
Mag	Strategy-defined qualitative effect label. With the built-in Mann-Whitney tests this is Cliff's delta classified by Romano (2006): neg (abs(δ) < 0.147), sml (< 0.33), med (< 0.474), lrg (≥ 0.474). For lrg, the cell is bold-red when the candidate is slower and bold-green when faster. - for the baseline or when significance is not tested. See Cliff's delta.
Alloc/op	Mean heap bytes per iteration (only visible when allocation tracking is enabled).

An optional Description column appears if any benchmark has a Description set.

For parameterized benchmarks, one column per parameter appears immediately after Benchmark, and the Benchmark cell shows the base method name. To save width, parametric tables label the comparison columns Ratio, Sig and Mag. When a single method is swept across parameter values, Ratio reports each point's scaling factor relative to the fastest point (the reference, shown as baseline); Sig and Mag stay -, because the engine does not test different workloads against one another. When a parameter group instead holds competing benchmarks, Sig and Mag carry the usual within-group significance and effect.

In Standard mode (--detail standard or WithDetail(ReportDetail.Standard)), the full multi-section output is shown: comparison table, Precision & Tail Latency, auto-tune, and Interpretation.

In Advanced mode (--detail advanced or WithDetail(ReportDetail.Advanced)), each benchmark row is followed by an indented stats block.

Adding progress display

ConsoleBenchmarkProgress displays warmup and measurement progress for each benchmark as it runs. It is independent of ConsoleReporter and can be used without it.

using NBenchmark.Reporters.Console;

await new BenchmarkSuite("name")
    .WithWarmup(25)        // pin so the progress bar has an exact total
    .WithIterations(200)   // pin so the progress bar has an exact total
    .WithReporter(new ConsoleReporter())
    .WithProgress(new ConsoleBenchmarkProgress())
    .RunAsync();

Pinning the warmup and sample counts gives the progress bar an exact total to track. With the default auto-resolved counts the bar fills toward the MaxSamples ceiling and the run usually stops earlier, once the confidence interval is tight enough.

Progress output is a live, updating line per benchmark:

──────────────── Running 2 benchmark(s) ────────────────

  [1/2] Compute ████████████░░░░░░░░ 60% measuring (120/200) ETA 0.4s
  ✓ Compute 12.4 ns (0.8s)
  ✓ Baseline 41.9 ns (1.1s)

──────────────── Completed in 1.9s ────────────────

Using with Benchmark (Single mode)

using NBenchmark.Reporters.Console;

var result = Benchmark.Run(() => MyMethod());
await result.PrintAsync();

PrintAsync runs the single result through ConsoleReporter and renders a table.

Printing markup from the summary line

The summary line at the bottom always shows the confidence level from the first successful result. If all benchmarks errored, only a list of error messages is shown.