Benchmarking .NET 6 applications using BenchmarkDotNet
Benchmarking is essential to the software development process, enabling developers to measure their applications’ performance and identify improvement areas. To accurately gauge the performance of .NET 6 applications, developers can rely on BenchmarkDotNet, a popular open-source benchmarking library.
Setting up BenchmarkDotNet
To install the BenchmarkDotNet NuGet package in our project, we run the following command in the project directory:
dotnet add package BenchmarkDotNet
Creating a benchmark
We will define a benchmark class. That contains the methods we want to benchmark. Check out the following example.
public class MathBenchmark{[Benchmark]public int SquareRootBenchmark(){return (int)Math.Sqrt(42);}[Benchmark]public int PowerOfTwoBenchmark(){return (int)Math.Pow(2, 10);}}
In the code above, we have defined a class named MathBenchmark, which contains two benchmark methods for measuring the performance of mathematical operations: SquareRootBenchmark() and PowerOfTwoBenchmark() methods. The [Benchmark] attribute is applied to these two methods. These methods perform square root and power-of-two calculations, respectively, and are intended to be used with the BenchmarkDotNet library to measure their execution times.
Running the benchmark
To execute the benchmark, our main function is as follows:
using BenchmarkDotNet.Attributes;
public class MathBenchmark
{
[Benchmark]
public float SquareRootBenchmark()
{
// return squareroot of first 1000 integers
float sum = 0;
for (int i = 0; i < 1000; i++)
{
sum += (int)Math.Sqrt(i);
}
return sum;
}
[Benchmark]
public int PowerOfTwoBenchmark()
{
// Return the sum of powers of two for the first 1000 integers
int sum = 0;
for (int i = 0; i < 1000; i++)
{
sum += (int)Math.Pow(2, i);
}
return sum;
}
}Note: When we run the application,
BenchmarkDotNetwill automatically detect the benchmark class and execute the defined benchmark methods.
In the MathBenchmark.cs file:
Line 1: We use the
BenchmarkDotNet.Attributesnamespace for benchmarking annotations and attributes.Line 3: We define a class named
MathBenchmarkwith benchmarking methods.Lines 5 and 17: We apply the
[Benchmark]attribute toSquareRootBenchmark()andPowerOfTwoBenchmark()methods.Lines 6–15: The
SquareRootBenchmark()method calculates the sum of square roots for the first 1000 integers.Lines 18–27: The
PowerOfTwoBenchmark()method calculates the sum of powers of two for the first 1000 integers.
In the Program.cs file:
Line 1: We use the
BenchmarkDotNet.Runningnamespace for benchmark execution.Line 3: We define an internal class named
Program.Lines 5–8: The entry point of the program is the
Main()method. It invokesBenchmarkRunner.Run<MathBenchmark>()and executes and runs the benchmarks defined in theMathBenchmarkclass. It captures and stores the benchmarking results in thesummaryvariable.
Analyzing benchmark results
BenchmarkDotNet generates detailed reports and statistical data to help us analyze the benchmark results effectively. By default, BenchmarkDotNet outputs its results in the console. However, we can also generate HTML or CSV reports for further analysis.
| Method | Mean | Error | StdDev ||-------------------- |----------:|----------:|----------:|| SquareRootBenchmark | 2.270 us | 0.0114 us | 0.0095 us || PowerOfTwoBenchmark | 24.471 us | 0.0541 us | 0.0422 us |
By following the steps above, we can have detailed performance measurements and analysis of our methods.
