Sustainable Code with .NET 9
Sustainable Code is a constantly growing GitHub repository created by me, in which I collect various everyday code snippets and measure the performance of the different implementation ways.
The goal is to create a collection of code that virtually everyone has in front of them every day and can thus easily implement the best way for themselves and their use case.
I took the current release of .NET 9 as an opportunity to update all the examples accordingly and compare them between .NET 7, .NET 8 and - if the dependencies have already been updated - .NET 9.
Here are my rough findings of some of the comparison snippets.
Condition vs Exception 📊
The goal here is to show why exceptions are not a suitable way to control application logic - and that it is advisable to avoid exceptions as much as possible, as recommended by the C# Coding Guidelines.
BenchmarkDotNet v0.14.0, Windows 10 (10.0.19045.5131/22H2/2022Update)
AMD Ryzen 9 9950X, 1 CPU, 32 logical and 16 physical cores
.NET SDK 9.0.100
[Host] : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 7.0 : .NET 7.0.20 (7.0.2024.26716), X64 RyuJIT AVX2
.NET 8.0 : .NET 8.0.11 (8.0.1124.51707), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 9.0 : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
| Method | Runtime | Mean | Error | StdDev | Ratio | RatioSD | Gen0 | Allocated |
|---------- |--------- |-------------:|----------:|----------:|---------:|--------:|-------:|----------:|
| Condition | .NET 7.0 | 11.495 ns | 0.0733 ns | 0.0612 ns | 10.30 | 0.06 | - | - |
| Condition | .NET 8.0 | 1.299 ns | 0.0041 ns | 0.0034 ns | 1.16 | 0.00 | - | - |
| Condition | .NET 9.0 | 1.116 ns | 0.0035 ns | 0.0031 ns | 1.00 | 0.00 | - | - |
| Exception | .NET 7.0 | 2,336.653 ns | 9.1249 ns | 8.5355 ns | 2,092.87 | 9.28 | 0.0114 | 232 B |
| Exception | .NET 8.0 | 2,303.988 ns | 2.6537 ns | 2.0719 ns | 2,063.61 | 5.80 | 0.0114 | 232 B |
| Exception | .NET 9.0 | 1,072.247 ns | 8.2707 ns | 7.3317 ns | 960.38 | 6.84 | 0.0134 | 224 B |
We can see here that exceptions have become significantly more efficient with .NET 9 - but we are still talking about thousand times slower behavior compared to a condition.
Full sample here.
Any vs Count > 0 📊
This example is about showing how Count vs. Any() works in terms of performance.
BenchmarkDotNet v0.14.0, Windows 10 (10.0.19045.5131/22H2/2022Update)
AMD Ryzen 9 9950X, 1 CPU, 32 logical and 16 physical cores
.NET SDK 9.0.100
[Host] : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 7.0 : .NET 7.0.20 (7.0.2024.26716), X64 RyuJIT AVX2
.NET 8.0 : .NET 8.0.11 (8.0.1124.51707), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 9.0 : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
| Method | Runtime | Items | Mean | StdDev | Ratio | RatioSD |
|------- |--------- |------ |----------:|----------:|------:|--------:|
| Any | .NET 7.0 | 0 | 1.9150 ns | 0.0271 ns | 4.069 | 0.06 |
| Any | .NET 8.0 | 0 | 1.8727 ns | 0.0139 ns | 3.979 | 0.04 |
| Any | .NET 9.0 | 0 | 0.4706 ns | 0.0031 ns | 1.000 | 0.01 |
| | | | | | | |
| Count | .NET 7.0 | 0 | 0.0034 ns | 0.0017 ns | 0.007 | 0.00 |
| Count | .NET 8.0 | 0 | 0.0038 ns | 0.0018 ns | 0.008 | 0.00 |
| Count | .NET 9.0 | 0 | 0.0161 ns | 0.0019 ns | 0.034 | 0.00 |
| | | | | | | |
| Any | .NET 7.0 | 1 | 1.9256 ns | 0.0194 ns | 4.06 | 0.05 |
| Any | .NET 8.0 | 1 | 1.8670 ns | 0.0057 ns | 3.94 | 0.02 |
| Any | .NET 9.0 | 1 | 0.4740 ns | 0.0027 ns | 1.00 | 0.01 |
| | | | | | | |
| Count | .NET 7.0 | 1 | 0.0082 ns | 0.0029 ns | 0.02 | 0.01 |
| Count | .NET 8.0 | 1 | 0.0062 ns | 0.0013 ns | 0.01 | 0.00 |
| Count | .NET 9.0 | 1 | 0.0189 ns | 0.0051 ns | 0.04 | 0.01 |
| | | | | | | |
| Any | .NET 7.0 | 10 | 1.9292 ns | 0.0199 ns | 4.09 | 0.04 |
| Any | .NET 8.0 | 10 | 1.8612 ns | 0.0050 ns | 3.95 | 0.02 |
| Any | .NET 9.0 | 10 | 0.4717 ns | 0.0015 ns | 1.00 | 0.00 |
| | | | | | | |
| Count | .NET 7.0 | 10 | 0.0107 ns | 0.0051 ns | 0.02 | 0.01 |
| Count | .NET 8.0 | 10 | 0.0056 ns | 0.0035 ns | 0.01 | 0.01 |
| Count | .NET 9.0 | 10 | 0.0159 ns | 0.0026 ns | 0.03 | 0.01 |
| | | | | | | |
| Any | .NET 8.0 | 100 | 1.8619 ns | 0.0126 ns | 3.93 | 0.05 |
| Any | .NET 7.0 | 100 | 1.9241 ns | 0.0252 ns | 4.06 | 0.07 |
| Any | .NET 9.0 | 100 | 0.4740 ns | 0.0055 ns | 1.00 | 0.02 |
| | | | | | | |
| Count | .NET 7.0 | 100 | 0.0102 ns | 0.0029 ns | 0.02 | 0.01 |
| Count | .NET 8.0 | 100 | 0.0058 ns | 0.0071 ns | 0.01 | 0.01 |
| Count | .NET 9.0 | 100 | 0.0157 ns | 0.0026 ns | 0.03 | 0.01 |
Count is still the much faster variant, but with .NET 9 there was a real performance boost for Linq.
Full sample here.
Small Hotpath Methods 📊
In the hotpath comparison - I am a fan of local methods and their better readability and performance benefit - there was no relevant improvement with .NET 9.
BenchmarkDotNet v0.14.0, Windows 10 (10.0.19045.5131/22H2/2022Update)
AMD Ryzen 9 9950X, 1 CPU, 32 logical and 16 physical cores
.NET SDK 9.0.100
[Host] : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 7.0 : .NET 7.0.20 (7.0.2024.26716), X64 RyuJIT AVX2
.NET 8.0 : .NET 8.0.11 (8.0.1124.51707), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 9.0 : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
| Method | Runtime | Iterations | Mean | Error | StdDev | Ratio |
|------------- |--------- |----------- |-------------:|-----------:|-----------:|------:|
| WithoutLocal | .NET 7.0 | 10 | 11.076 ns | 0.1264 ns | 0.1121 ns | 3.94 |
| WithoutLocal | .NET 8.0 | 10 | 2.791 ns | 0.0315 ns | 0.0295 ns | 0.99 |
| WithoutLocal | .NET 9.0 | 10 | 2.812 ns | 0.0181 ns | 0.0151 ns | 1.00 |
| | | | | | | |
| WithLocal | .NET 7.0 | 10 | 3.494 ns | 0.0859 ns | 0.0990 ns | 1.24 |
| WithLocal | .NET 8.0 | 10 | 2.807 ns | 0.0147 ns | 0.0130 ns | 0.99 |
| WithLocal | .NET 9.0 | 10 | 2.827 ns | 0.0277 ns | 0.0246 ns | 1.00 |
| | | | | | | |
| WithoutLocal | .NET 7.0 | 100 | 119.236 ns | 1.5937 ns | 1.4907 ns | 4.91 |
| WithoutLocal | .NET 8.0 | 100 | 25.734 ns | 0.1918 ns | 0.1794 ns | 1.06 |
| WithoutLocal | .NET 9.0 | 100 | 24.262 ns | 0.2096 ns | 0.1960 ns | 1.00 |
| | | | | | | |
| WithLocal | .NET 7.0 | 100 | 40.118 ns | 0.8185 ns | 0.8758 ns | 1.64 |
| WithLocal | .NET 8.0 | 100 | 25.077 ns | 0.3515 ns | 0.3288 ns | 1.03 |
| WithLocal | .NET 9.0 | 100 | 24.466 ns | 0.2786 ns | 0.2606 ns | 1.00 |
| | | | | | | |
| WithoutLocal | .NET 7.0 | 1000 | 1,135.122 ns | 17.3263 ns | 16.2071 ns | 4.74 |
| WithoutLocal | .NET 8.0 | 1000 | 250.022 ns | 1.5989 ns | 1.4174 ns | 1.04 |
| WithoutLocal | .NET 9.0 | 1000 | 239.307 ns | 1.5506 ns | 1.4504 ns | 1.00 |
| | | | | | | |
| WithLocal | .NET 7.0 | 1000 | 330.232 ns | 5.3309 ns | 4.9866 ns | 1.37 |
| WithLocal | .NET 8.0 | 1000 | 241.045 ns | 2.0723 ns | 1.9384 ns | 1.00 |
| WithLocal | .NET 9.0 | 1000 | 241.217 ns | 1.8586 ns | 1.6476 ns | 1.00 |
Full sample here.
Source Generator based Logging Messages 📊
Logging is expensive, so it is advisable to use compile-time logging through Source Code Generators.
BenchmarkDotNet v0.14.0, Windows 10 (10.0.19045.5131/22H2/2022Update)
AMD Ryzen 9 9950X, 1 CPU, 32 logical and 16 physical cores
.NET SDK 9.0.100
[Host] : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 7.0 : .NET 7.0.20 (7.0.2024.26716), X64 RyuJIT AVX2
.NET 8.0 : .NET 8.0.11 (8.0.1124.51707), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
.NET 9.0 : .NET 9.0.0 (9.0.24.52809), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
| Method | Runtime | Mean | Error | StdDev | Ratio | RatioSD | Gen0 | Allocated |
|-------------------- |--------- |----------:|---------:|---------:|------:|--------:|-------:|----------:|
| SourceCodeGenerated | .NET 7.0 | 83.49 ns | 0.299 ns | 0.279 ns | 1.24 | 0.01 | - | - |
| SourceCodeGenerated | .NET 8.0 | 82.12 ns | 0.252 ns | 0.224 ns | 1.22 | 0.01 | - | - |
| SourceCodeGenerated | .NET 9.0 | 67.30 ns | 0.365 ns | 0.341 ns | 1.00 | 0.01 | - | - |
| | | | | | | | | |
| Concat | .NET 7.0 | 112.57 ns | 0.235 ns | 0.208 ns | 1.67 | 0.01 | 0.0286 | 480 B |
| Concat | .NET 8.0 | 96.94 ns | 1.239 ns | 1.098 ns | 1.44 | 0.02 | 0.0286 | 480 B |
| Concat | .NET 9.0 | 94.59 ns | 1.136 ns | 1.007 ns | 1.41 | 0.02 | 0.0286 | 480 B |
| | | | | | | | | |
| Interpolation | .NET 7.0 | 112.79 ns | 0.444 ns | 0.416 ns | 1.68 | 0.01 | 0.0286 | 480 B |
| Interpolation | .NET 8.0 | 97.08 ns | 1.208 ns | 1.071 ns | 1.44 | 0.02 | 0.0286 | 480 B |
| Interpolation | .NET 9.0 | 94.12 ns | 0.620 ns | 0.580 ns | 1.40 | 0.01 | 0.0286 | 480 B |
With .NET 9, this recommended variant has become even faster, making implementation all the more worthwhile, both in terms of memory allocation and performance.
Full sample here.
Summary
.NET 9 is another release that brings with it an enormous performance benefit and thus makes an entire software industry more efficient - for those who rely on .NET.
Autor
Benjamin Abt
Ben is a passionate developer and software architect and especially focused on .NET, cloud and IoT. In his professional he works on high-scalable platforms for IoT and Industry 4.0 focused on the next generation of connected industry based on Azure and .NET. He runs the largest german-speaking C# forum myCSharp.de, is the founder of the Azure UserGroup Stuttgart, a co-organizer of the AzureSaturday, runs his blog, participates in open source projects, speaks at various conferences and user groups and also has a bit free time. He is a Microsoft MVP since 2015 for .NET and Azure.