# How has LINQ performance enhanced in \.NET 7? New version of \.NET enhanced the performance of the Min, Max, Average and Sum methods for arrays and lists\. How much do you think their execution speed has increased? 2x or 5x? No, they got even faster\. Let's see how that was achieved\. ![1011_LINQ_Performance_dotNET7/image1.png](https://import.viva64.com/docx/blog/1011_LINQ_Performance_dotNET7/image1.png) ## How has LINQ enhanced? LINQ \(Language\-Integrated Query\) is a simple and convenient query language\. It allows you to express complex operations in a simple way\. Almost every \.NET developer uses LINQ\. However, this simplicity of use comes at a price of the execution speed and extra memory allocation\. In most situations, it has no significant effect\. However, in cases when performance is critical, these limitations may be pretty unpleasant\. So, the recent update has enhanced the performance of the following methods: _Enumerable\.Max_, _Enumerable\.Min_, _Enumerable\.Average_, _Enumerable\.Sum_\. Let's see how their performance increased using the following benchmark: ```cpp using BenchmarkDotNet.Attributes; using BenchmarkDotNet.Running; using System.Collections.Generic; using System.Linq; [MemoryDiagnoser(displayGenColumns: false)] public partial class Program { static void Main(string[] args) => BenchmarkSwitcher.FromAssembly(typeof(Program).Assembly).Run(args); [Params (10, 10000)] public int Size { get; set; } private IEnumerable items; [GlobalSetup] public void Setup() { items = Enumerable.Range(1, Size).ToArray(); } [Benchmark] public int Min() => items.Min(); [Benchmark] public int Max() => items.Max(); [Benchmark] public double Average() => items.Average(); [Benchmark] public int Sum() => items.Sum(); } ``` Benchmark results: |Method|Runtime|Size|Mean|Ratio|Allocated| |---|---|---|---|---|---| |Min|\\\.NET 6\\\.0|10|75\\\.491 ns|1\\\.00|32 B| |Min|\\\.NET 7\\\.0|10|7\\\.749 ns|0\\\.10|\\\-| ||||||| |Max|\\\.NET 6\\\.0|10|71\\\.128 ns|1\\\.00|32 B| |Max|\\\.NET 7\\\.0|10|6\\\.493 ns|0\\\.09|\\\-| ||||||| |Average|\\\.NET 6\\\.0|10|68\\\.963 ns|1\\\.00|32 B| |Average|\\\.NET 7\\\.0|10|7\\\.315 ns|0\\\.11|\\\-| ||||||| |Sum|\\\.NET 6\\\.0|10|69\\\.509 ns|1\\\.00|32 B| |Sum|\\\.NET 7\\\.0|10|9\\\.058 ns|0\\\.13|\\\-| ||||||| |Min|\\\.NET 6\\\.0|10000|61,567\\\.392 ns|1\\\.00|32 B| |Min|\\\.NET 7\\\.0|10000|2,967\\\.947 ns|0\\\.05|\\\-| ||||||| |Max|\\\.NET 6\\\.0|10000|56,106\\\.592 ns|1\\\.00|32 B| |Max|\\\.NET 7\\\.0|10000|2,948\\\.302 ns|0\\\.05|\\\-| ||||||| |Average|\\\.NET 6\\\.0|10000|52,803\\\.907 ns|1\\\.00|32 B| |Average|\\\.NET 7\\\.0|10000|2,967\\\.810 ns|0\\\.06|\\\-| ||||||| |Sum|\\\.NET 6\\\.0|10000|52,732\\\.121 ns|1\\\.00|32 B| |Sum|\\\.NET 7\\\.0|10000|5,897\\\.220 ns|0\\\.11|\\\-| The results show that the execution time for finding the minimum element of an array has generally decreased\. In 10 times for small arrays and 20 times for arrays containing 10,000 elements\. Similarly, for other methods \(except for finding the sum, the difference between the sizes of the collections did not affect the results much\)\. It is also worth noting that in \.NET 7 no additional memory is allocated when methods are called\. Let's check how these methods work with _List_\. |Method|Runtime|Size|Mean|Ratio|Allocated| |---|---|---|---|---|---| |Min|\\\.NET 6\\\.0|10|122\\\.554 ns|1\\\.00|40 B| |Min|\\\.NET 7\\\.0|10|8\\\.995 ns|0\\\.07|\\\-| ||||||| |Max|\\\.NET 6\\\.0|10|115\\\.135 ns|1\\\.00|40 B| |Max|\\\.NET 7\\\.0|10|9\\\.171 ns|0\\\.08|\\\-| ||||||| |Average|\\\.NET 6\\\.0|10|110\\\.825 ns|1\\\.00|40 B| |Average|\\\.NET 7\\\.0|10|8\\\.163 ns|0\\\.07|\\\-| ||||||| |Sum|\\\.NET 6\\\.0|10|113\\\.812 ns|1\\\.00|40 B| |Sum|\\\.NET 7\\\.0|10|13\\\.197 ns|0\\\.12|\\\-| ||||||| |Min|\\\.NET 6\\\.0|10000|91,529\\\.841 ns|1\\\.00|40 B| |Min|\\\.NET 7\\\.0|10000|2,941\\\.226 ns|0\\\.03|\\\-| ||||||| |Max|\\\.NET 6\\\.0|10000|84,565\\\.787 ns|1\\\.00|40 B| |Max|\\\.NET 7\\\.0|10000|2,957\\\.451 ns|0\\\.03|\\\-| ||||||| |Average|\\\.NET 6\\\.0|10000|81,205\\\.103 ns|1\\\.00|40 B| |Average|\\\.NET 7\\\.0|10000|2,959\\\.882 ns|0\\\.04|\\\-| ||||||| |Sum|\\\.NET 6\\\.0|10000|81,857\\\.576 ns|1\\\.00|40 B| |Sum|\\\.NET 7\\\.0|10000|5,783\\\.370 ns|0\\\.07|\\\-| In \.NET 6, all operations on arrays much faster than on lists\. The same is true for small collections in \.NET 7\. As the number of elements increases, lists are equal to arrays in performance\. According to the test results, lists' performance increased by 31 times\. ## But how could this be achieved? Let's take a closer look at the implementation of the _Min_ method\. That's how the _Min_ method is implemented in \.NET 6: ```cpp public static int Min(this IEnumerable source) { if (source == null) { ThrowHelper.ThrowArgumentNullException(ExceptionArgument.source); } int value; using (IEnumerator e = source.GetEnumerator()) { if (!e.MoveNext()) { ThrowHelper.ThrowNoElementsException(); } value = e.Current; while (e.MoveNext()) { int x = e.Current; if (x < value) { value = x; } } } return value; } ``` The method is quite simple\. We get the _IEnumerable_ collection, take the collection element and use the _MoveNext_ method to get the next element\. We compare them, save the one that is smaller, and repeat until the end of the collection\. The new version of the _Min_ method is different: ```cpp public static int Min(this IEnumerable source) => MinInteger(source); ``` The _MinInteger_ method is applied to a collection of integers\. Let's examine it in more details\. ```cpp private static T MinInteger(this IEnumerable source) where T : struct, IBinaryInteger { T value; if (source.TryGetSpan(out ReadOnlySpan span)) { if (Vector.IsHardwareAccelerated && span.Length >= Vector.Count * 2) { .... // Optimized implementation return ....; } } .... //Implementation as in .NET 6 } ``` Primarily, we try to get the object of the _ReadOnlySpan_ type from the provided collection\. If we fail to get the _ReadOnlySpan_ collection representation, then the code branch \(that matches the implementation of the _Min_ method in \.NET 6\) is executed\. In our case, we can get _ReadOnlySpan_, since we use arrays and lists\. But what is _ReadOnlySpan_ actually? The _Span_ and _ReadOnlySpan_ types provide safe representation of a continuous managed and unmanaged memory area\. The structure of the _Span_ type is defined as a _ref struct\._ This means that it can only be placed on the stack, which makes it possible to avoid allocating additional memory and improves data performance\. The_ Span_ type has also undergone some changes in the new version of C\#\. Since C\# 11 introduced the option to create reference fields in a _ref struct_, the internal representation of _Span_ has changed\. Previously, a special internal type — _ByReference_, was used to store a reference to the beginning of the memory area, but there was no security check in it\. Now _ref fields_ are used for this purpose\. It provides a more secure memory operation\. You can read more about other changes in C\# 11 in the [article](https://pvs-studio.com/en/blog/posts/csharp/1002/)\. But let's get back to the _Min_ method\. When you get _ReadOnlySpan_, the method tries to perform a vector search using the _Vector_ type\. To do this, the following condition should be met: ```cpp if (Vector.IsHardwareAccelerated && span.Length >= Vector.Count * 2) ``` The first part of the condition checks whether the Vector\.IsHardwareAccelerated property returns true\. Let's take a look at the implementation of this property\. ```cpp public static bool IsHardwareAccelerated { [Intrinsic] get => false; } ``` The _\[Intrinsic\]_ attribute is applied to the getter\. The attribute indicates that the value returned by _IsHardwareAccelerated_ can be replaced JIT\. The property returns _true_ if hardware acceleration can be applied to operations on vectors through built\-in JIT support, otherwise _false_ is returned\. To enable hardware acceleration, you need to run the build for the x64 platform with the Release configuration or build the project for AnyCPU with the "Prefer 32\-bit" setting disabled\. To fulfill the second part of the condition, the size of _Span_ should be at least 2 times the size of the vector\. How is the size of this vector calculated? The _Vector_ data type is used in the new implementation of the _Min_ method to create a vector\. This type is a wrapper for three other types: _Vector64_, _Vector128_ and _Vector256_\. They contain vectorized data of the corresponding length\. _Vector128_ can store 16 8\-bit, 8 16\-bit, 4 32\-bit or 2 64\-bit values\. The type to be used is selected depending on whether it is supported by the processor or not\. Thus, if the _Vector128_ type is used when executing the method, then the _Span_ type obtained from the array or list must contain 8 or more elements of the _int_ type\. If all conditions are met, the method uses the advantages of the _ReadOnlySpan_ and _Vector_ types to optimize finding the minimum: ```cpp private static T MinInteger(this IEnumerable source) where T : struct, IBinaryInteger { .... if (Vector.IsHardwareAccelerated && span.Length >= Vector.Count * 2) { var mins = new Vector(span); index = Vector.Count; do { mins = Vector.Min(mins, new Vector(span.Slice(index))); index += Vector.Count; } while (index + Vector.Count <= span.Length); value = mins[0]; for (int i = 1; i < Vector.Count; i++) { if (mins[i] < value) { value = mins[i]; } } .... } ``` Let's explore how the vectorized implementation of finding a minimum works\. A vector is created from the first N _Span_ elements \(N depends on the type of vector; for _Vector128_ it is 4 elements\)\. This vector is compared with a vector of the following N elements in _Vector\.Min_\. The resulting vector contains the minimum value of each pair of elements of the two given vectors\. Then the next part of _Span_ is taken, and the search continues\. You only need to find the minimum value among those stored in the resulting vector\. The advantage of using a vector is that all operations on its elements occur simultaneously\. An example of using _Span_ and vectorization to optimize the _Min_ method is shown above\. But what about the other methods? You can meet similar features for the _Max_, _Average_, _Sum_ methods\. The optimized versions of these methods are available for arrays and lists of _int_, _long_, _float_, _double_ and _decimal_ types\. ## Conclusion Microsoft developers used _Span_ and hardware acceleration to work with vectors\. As a result, the \.NET community brought visible enhancement to LINQ performance\. Now we can use advanced methods in cases where the speed of code execution, and allocated memory are critical\. Actually, \.NET 7 got a few other performance improvements\. You can read about them in [\.NET Blog](https://devblogs.microsoft.com/dotnet/performance_improvements_in_net_7/)\.