Developers often point to architecture, language choice, or hardware when trying to explain why an application is slow — but code-level optimizations are frequently the first place to look. Before optimizing, you need to measure. Would 1 ms make a difference? In most production systems the answer is obviously yes.

BenchmarkDotNet

BenchmarkDotNet is the standard .NET benchmarking library. Install it from NuGet:

NuGet: dotnet add package BenchmarkDotNet

Setup

Decorate your benchmark class with [MemoryDiagnoser] to track allocations and [MediumRunJob] to control the iteration count. Mark the baseline method with [Benchmark(Baseline = true)] and use [ArgumentsSource] to supply parameterized inputs from a method:

csharp

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;
using BenchmarkDotNet.Jobs;
using BenchmarkDotNet.Columns;

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

[MemoryDiagnoser]
[MediumRunJob, SkewnessColumn, KurtosisColumn]
public class ListHelper
{
    public IEnumerable<string> Password()
    {
        yield return string.Join(',', Enumerable.Range(1, 31).Select(n => n.ToString()));
    }

    [Benchmark(Baseline = true)]
    [ArgumentsSource(nameof(Password))]
    public byte[] HashPwd(string password)
    {
        byte[] salt = new byte[32];
        RNGCryptoServiceProvider.Create().GetBytes(salt);

        byte[] plainTextBytes = System.Text.Encoding.Unicode.GetBytes(password);
        byte[] combinedBytes   = new byte[plainTextBytes.Length + salt.Length];
        Buffer.BlockCopy(plainTextBytes, 0, combinedBytes, 0, plainTextBytes.Length);
        Buffer.BlockCopy(salt, 0, combinedBytes, plainTextBytes.Length, salt.Length);

        HashAlgorithm hashAlgo = new SHA256Managed();
        byte[] hash             = hashAlgo.ComputeHash(combinedBytes);

        byte[] hashPlusSalt = new byte[hash.Length + salt.Length];
        Buffer.BlockCopy(hash, 0, hashPlusSalt, 0, hash.Length);
        Buffer.BlockCopy(salt, 0, hashPlusSalt, hash.Length, salt.Length);

        return hashPlusSalt;
    }
}

Reading the Results

BenchmarkDotNet prints a results table with columns like Mean, Error, StdDev, Ratio, and memory columns Gen0 / Allocated. The Ratio column compares each method against the baseline. Skewness and kurtosis columns (from the SkewnessColumn / KurtosisColumn attributes) indicate distribution shape — high skewness can reveal occasional slow outliers.

Always run benchmarks in Release mode. BenchmarkDotNet enforces this and warns if you run from the debugger.

Net Core Benchmark

BenchmarkDotNet

Setup

Reading the Results