Regex benchmarks — gdhami.net

BENCHMARKS · REGEX

One pattern, three engines: interpreted, compiled, source-generated. Suite and raw reports: github.com/gdhami-net/dotnet-benchmarks.

Intel Core Ultra 9 285HX · Windows 11 · BenchmarkDotNet 0.15.8 · SDK 10.0.201 · 2026-09-01 · median of 4 quiet runs · typical spread ±3.3% · bars/tables: mean time, lower is better

EmailMatching

The same email pattern over 5,000 mixed lines: interpreted vs Compiled vs source-generated regex.

TAKEAWAY A static [GeneratedRegex] is the right default: compiled-class speed, no startup JIT cost, AOT-compatible. Never new up a Regex per call.
In this run: SourceGen_regex is fastest — 1.8× faster than the baseline.

RESULTS

method	net9.0	net10.0	ratio	allocated	Δ net10.0
Interpreted_regex baseline	254.2 µs ±920 ns	248.7 µs ±1.4 µs	1.00×	—	-2%
Compiled_regex	152.3 µs ±65 ns	147.1 µs ±92 ns	0.59×	—	-3%
SourceGen_regex	122.9 µs ±99 ns	140.1 µs ±87 ns	0.56×	—	+14%

RATIO VS BASELINE · net10.0

Interpreted_regex1.00×

Compiled_regex0.59×

SourceGen_regex0.56×

THE LEDGER · SAME WORKLOAD ACROSS RELEASES (LOG)

Interpreted_regexCompiled_regexSourceGen_regex

THE CODE BEING MEASURED

Interpreted_regex — what this measures

new Regex(pattern) — interpreted at match time.

[Benchmark (Baseline)]
public int Interpreted_regex()
{
    var hits = 0;
    foreach (var l in _lines)
        if (Interpreted.IsMatch(l)) hits++;
    return hits;
}

Compiled_regex — what this measures

RegexOptions.Compiled — IL emitted at construction.

[Benchmark]
public int Compiled_regex()
{
    var hits = 0;
    foreach (var l in _lines)
        if (Compiled.IsMatch(l)) hits++;
    return hits;
}

SourceGen_regex — what this measures

[GeneratedRegex] — C# generated at compile time, AOT-friendly.

[Benchmark]
public int SourceGen_regex()
{
    var hits = 0;
    foreach (var l in _lines)
        if (SourceGen().IsMatch(l)) hits++;
    return hits;
}