Benchmarking .NET 5 code with BenchmarkDotNet

If you need to measure performance changes in .NET 5 code, one of the easiest ways is to use a new tool called BenchmarkDotNet. It is a benchmarking library for .NET applications that handles warmup runs, iterations, statistics, and many of the common mistakes found in hand-written benchmarks.

You can add the functionality as a NuGet package to your application. For example, using dotnet add package:

dotnet add package BenchmarkDotNet

Then create a simple benchmark class and apply the attributes:

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;

public class StringBenchmarks
{
    [Benchmark]
    public string StringConcat()
    {
        return "Hello" + " " + "world";
    }

    [Benchmark]
    public string StringInterpolation()
    {
        return $"{ "Hello" } { "world" }";
    }
}

public class Program
{
    public static void Main(string[] args)
    {
        BenchmarkRunner.Run<StringBenchmarks>();
    }
}

Run the benchmark in Release mode:

dotnet run -c Release

The tool creates a detailed summary including execution time, standard deviation, allocation data, and comparisons between methods.

Tip: always benchmark Release builds. BenchmarkDotNet warns about Debug builds because the results are usually unreliable.

A useful feature is the memory diagnoser, which helps detect unnecessary allocations:

[MemoryDiagnoser]
public class StringBenchmarks
{
    // put your benchmarks here
}

This adds allocation statistics directly to the benchmark results.

BenchmarkDotNet also supports comparing multiple runtimes. During the .NET 5 timeframe, this is handy for checking differences between .NET Core 3.1 and .NET 5, for instance. Something along these lines:

[SimpleJob(RuntimeMoniker.NetCoreApp31)]
[SimpleJob(RuntimeMoniker.Net50)]
public class StringBenchmarks
{
    // your benchmarks go here
}

For performance-sensitive ASP.NET Core or backend code, having repeatable benchmarks makes it much easier to validate whether a change actually improves throughput or just "feels" faster.

Happy hacking!