Avoid sort allocations #116109

2A5F · 2025-05-29T18:39:24Z

Modified ArraySortHelper and added ArraySortHelper<T, TComparer> to allow generic struct TComparer without memory allocation

Span.Sort<T, TComparer> will not box now

#39466
#39543

jkotas · 2025-05-31T05:24:52Z

@EgorBot -intel

using BenchmarkDotNet.Attributes;

public class Bench
{
    int[] a = new int[10000];
    Comparison<int> c = (int x, int y) => x - y;

    [IterationSetup]
    public void IterationSetup() { for (int i = 0; i < a.Length; i++) a[i] = i; }

    [Benchmark]
    public void Sort() => Array.Sort(a, c);
}

jkotas · 2025-05-31T06:04:51Z

Could you please share perf numbers that demonstrate the performance improvements?

The one micro-benchmark that I have kicked off above shows ~10% regression: EgorBot/runtime-utils#370 (comment) .

2A5F · 2025-05-31T08:25:07Z

No, I haven't done any performance testing.
Maybe passing TComparer by ref causes performance overhead.
Can this test bot output jit diffs?

adamsitnik · 2025-06-02T08:37:45Z

@2A5F You can use the benchmarks from the performance repo (we use them for ensuring there are no regressions):

https://github.com/dotnet/performance/blob/3499787dbbde3807402516462453e67097066d4a/src/benchmarks/micro/libraries/System.Collections/Sort.cs#L53-L54

You can both trace files and disassembler output:

https://github.com/dotnet/performance/tree/main/src/benchmarks/micro#quick-start

2A5F · 2025-06-02T15:30:58Z

@adamsitnik Thanks for the notice
@jkotas After testing, I think it is impossible to unify the generic paths without regression, so for now I will only use generic in the Span.Sort<TComparer>.

Test Code

base on https://github.com/dotnet/performance/blob/3499787dbbde3807402516462453e67097066d4a/src/benchmarks/micro/libraries/System.Collections/Sort.cs

[Benchmark]
public void Span_ComparerStruct() => _arrays[_iterationIndex++].AsSpan(0, Size).Sort(new ComparableComparerStruct());

private readonly struct ComparableComparerStruct : IComparer<T>
{
    [MethodImpl(MethodImplOptions.AggressiveInlining)]
    public int Compare(T x, T y) => x.CompareTo(y);
}

All test on

BenchmarkDotNet v0.14.1-nightly.20250107.205, Windows 10 (10.0.20348.1906) (VMware)
AMD Ryzen 9 7950X 4.49GHz, 16 CPU, 32 logical and 32 physical cores
.NET SDK 10.0.100-preview.6.25281.103
  [Host]     : .NET 10.0.0 (10.0.25.28203), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
  Job-VBHUQD : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI
  Job-BXQXFN : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX-512F+CD+BW+DQ+VL+VBMI

PowerPlanMode=00000000-0000-0000-0000-000000000000  InvocationCount=5000  IterationTime=250ms
MaxIterationCount=20  MinIterationCount=15  MinWarmupIterationCount=6
UnrollFactor=1  WarmupCount=-1

Int32

Method	Job	Toolchain	Size	Mean	Error	StdDev	Median	Min	Max	Ratio	RatioSD	Gen0	Code Size	Allocated	Alloc Ratio
Span_ComparerStruct	Job-VBHUQD	pr	512	3.562 μs	0.1579 μs	0.1819 μs	3.547 μs	3.312 μs	3.931 μs	0.30	0.02	-	1,126 B	-	0.00
Span_ComparerStruct	Job-BXQXFN	main	512	11.943 μs	0.2768 μs	0.3188 μs	11.825 μs	11.523 μs	12.474 μs	1.00	0.04	-	1,910 B	88 B	1.00

BigStruct

Method	Job	Toolchain	Size	Mean	Error	StdDev	Median	Min	Max	Ratio	RatioSD	Gen0	Code Size	Allocated	Alloc Ratio
Span_ComparerStruct	Job-VBHUQD	pr	512	4.677 μs	0.1230 μs	0.1417 μs	4.645 μs	4.362 μs	4.949 μs	0.34	0.01	-	1,352 B	-	0.00
Span_ComparerStruct	Job-BXQXFN	main	512	13.659 μs	0.2965 μs	0.3173 μs	13.594 μs	13.215 μs	14.288 μs	1.00	0.03	-	2,426 B	88 B	1.00

IntClass

Method	Job	Toolchain	Size	Mean	Error	StdDev	Median	Min	Max	Ratio	RatioSD	Gen0	Code Size	Allocated	Alloc Ratio
Span_ComparerStruct	Job-VBHUQD	pr	512	23.89 μs	0.541 μs	0.601 μs	23.92 μs	22.59 μs	24.86 μs	0.92	0.03	-	2,958 B	-	0.00
Span_ComparerStruct	Job-BXQXFN	main	512	26.00 μs	0.519 μs	0.556 μs	26.01 μs	24.97 μs	26.85 μs	1.00	0.03	-	968 B	88 B	1.00

IntStruct

Method	Job	Toolchain	Size	Mean	Error	StdDev	Median	Min	Max	Ratio	RatioSD	Gen0	Code Size	Allocated	Alloc Ratio
Span_ComparerStruct	Job-VBHUQD	pr	512	3.054 μs	0.1060 μs	0.1221 μs	3.006 μs	2.880 μs	3.293 μs	0.26	0.01	-	1,102 B	-	0.00
Span_ComparerStruct	Job-BXQXFN	main	512	11.784 μs	0.2308 μs	0.2159 μs	11.787 μs	11.312 μs	12.174 μs	1.00	0.03	-	1,916 B	88 B	1.00

String

Method	Job	Toolchain	Size	Mean	Error	StdDev	Median	Min	Max	Ratio	RatioSD	Gen0	Code Size	Allocated	Alloc Ratio
Span_ComparerStruct	Job-VBHUQD	pr	512	153.6 μs	1.99 μs	1.76 μs	153.6 μs	150.0 μs	156.8 μs	1.00	0.02	-	2,951 B	-	0.00
Span_ComparerStruct	Job-BXQXFN	main	512	154.3 μs	2.19 μs	2.05 μs	153.6 μs	151.8 μs	158.0 μs	1.00	0.02	-	968 B	88 B	1.00

2A5F · 2025-06-02T16:45:56Z

Ideas and todo:

Consider get a pointer to interface method to simulate a value type delegate
Delegate struct wrapper has 1% regression (this test result may be invalid), maybe this is acceptable, or try take out the pointer and rewrap it
Null ICompear<T> path can replace to Compear<T>.Default to avoid interface virtual call

Not feasible, C# does not have syntax to get instance method pointers. Even if use InlineIL.Fody to get it, the benchmark is not as good as the delegate.
Since 1 is not feasible, then 2 is also not feasible.

andrewjsaid · 2025-06-03T13:32:52Z

To avoid duplicate code, could you do something like this:

struct DelegateWrappedComparer<T> : IComparer<T>
{
    private Comparer<T> _comparer;

    public int Compare(T? x, T? y) => _comparer.Compare(x, y);
}

And point the "old" version of the method to the new one, with this wrapper?

Hope this is clear.

2A5F · 2025-06-03T13:48:19Z

@andrewjsaid For Comparer<T>, I'm not sure if this makes sense. Perhaps a struct wrapper can convert an interface virtual call into an abstract class virtual call; however public apis uses only IComparer<T> not Comparer<T>. And Comparer<T>.Default have some jit magic (I don't know what it is, but the test results are like this) so dose not need struct wrapper.

Delegate is Comparison<T> and the test results have regression EgorBot/runtime-utils#370 (comment)

This test has expired/is invalid. I will retest the new code later when I have time.

(benchmarks are really slow)

andrewjsaid · 2025-06-03T14:07:34Z

My reasoning was that by using the struct wrapper, methods like the below would be specialized for that struct and the JIT would be able to inline the call to DelegateWrappedComparer<T>.CompareTo thus making the SwapIfGreater<TComparer>(...) and SwapIfGreater(...)` equivalent.

private static void SwapIfGreater<TComparer>(Span<T> keys, TComparer comparer, int i, int j)
  where TComparer : IComparer<T>, allows ref struct

However on further reflection I note that currently since SwapIfGreater is not generic, by specializing it per DelegateWrappedComparer<T> this would mean that instead of having a single re-usable method if T is a class (via __CANON) the JIT would have to create a new copy for each T which can create significant bloat. Not sure if worth the maintenance cost of duplicate methods.

2A5F · 2025-06-03T14:16:47Z

However on further reflection I note that currently since SwapIfGreater is not generic, by specializing it per DelegateWrappedComparer this would mean that instead of having a single re-usable method if T is a class (via __CANON) the JIT would have to create a new copy for each T which can create significant bloat. Not sure if worth the maintenance cost of duplicate methods.

#39466 (comment)

andrewjsaid · 2025-06-03T14:19:55Z

Oops already suggested 🤦
Too many issues / PRs with similar names I missed that one

Thanks for the link

2A5F · 2025-09-25T04:37:39Z

I accidentally pushed to the wrong branch; the tags and review may need to be adjusted.

src/libraries/System.Private.CoreLib/src/System/Collections/Generic/ArraySortHelper.cs

src/libraries/System.Private.CoreLib/src/System/MemoryExtensions.cs

jkotas · 2025-09-25T04:47:44Z

Considering that, should we close this PR? or is there any suggestion to improve the regressed case?

The throughput should be fixable by introducing even more code duplication. (A separate question is whether the code duplication is worth it.)

src/libraries/System.Private.CoreLib/src/System/MemoryExtensions.cs

tarekgh · 2025-10-01T00:24:20Z

I assume this is related?

https://github.com/dotnet/runtime/pull/116109/checks?check_run_id=51639673842

System.InvalidOperationException : Failed to compare two elements in the array.
---- System.NullReferenceException : Object reference not set to an instance of an object

CallStack
   at System.Collections.Generic.ArraySortHelperForTComparer`3[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e],[System.Int32, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e],[System.Collections.Generic.IComparer`1[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]], System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]].Sort(Span`1 keys, Span`1 values, IComparer`1 comparer) in /_/src/libraries/System.Private.CoreLib/src/System/Collections/Generic/ArraySortHelper.cs:line 1367
   at System.MemoryExtensions.Sort[IntPtr,Int32,IComparer`1](Span`1 keys, Span`1 items, IComparer`1 comparer) in /_/src/libraries/System.Private.CoreLib/src/System/MemoryExtensions.cs:line 4393
   at System.MemoryExtensions.Sort[IntPtr,Int32](Span`1 keys, Span`1 items) in /_/src/libraries/System.Private.CoreLib/src/System/MemoryExtensions.cs:line 4363
   at System.Numerics.Tensors.TensorShape..ctor(IntPtr linearLength, ReadOnlySpan`1 lengths, ReadOnlySpan`1 strides, TensorFlags flags) in /_/src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/TensorShape.cs:line 180
   at System.Numerics.Tensors.TensorShape.Create[Int32](Int32[] array, Int32 start, ReadOnlySpan`1 lengths, ReadOnlySpan`1 strides) in /_/src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/TensorShape.cs:line 893
   at System.Numerics.Tensors.ReadOnlyTensorSpan`1[[System.Int32, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]]..ctor(Int32[] array, Int32 start, ReadOnlySpan`1 lengths, ReadOnlySpan`1 strides) in /_/src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/netcore/ReadOnlyTensorSpan_1.cs:line 102
   at System.Numerics.Tensors.Tests.ReadOnlyTensorSpanTests.ReadOnlyTensorSpanArrayConstructorTests() in /_/src/libraries/System.Numerics.Tensors/tests/ReadOnlyTensorSpanTests.cs:line 322
   at System.Reflection.MethodBaseInvoker.InterpretedInvoke_Method(Object obj, IntPtr* args) in /_/src/mono/System.Private.CoreLib/src/System/Reflection/MethodBaseInvoker.Mono.cs:line 22
   at System.Reflection.MethodBaseInvoker.InvokeWithNoArgs(Object obj, BindingFlags invokeAttr) in /_/src/libraries/System.Private.CoreLib/src/System/Reflection/MethodBaseInvoker.cs:line 57
----- Inner Stack Trace -----
   at System.Collections.Generic.ArraySortHelperForTComparer`3[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e],[System.Int32, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e],[System.Collections.Generic.IComparer`1[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]], System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]].SwapIfGreaterWithValues(Span`1 keys, Span`1 values, IComparer`1 comparer, Int32 i, Int32 j) in /_/src/libraries/System.Private.CoreLib/src/System/Collections/Generic/ArraySortHelper.cs:line 1392
   at System.Collections.Generic.ArraySortHelperForTComparer`3[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e],[System.Int32, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e],[System.Collections.Generic.IComparer`1[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]], System.Private.CoreLib, Version=10.0.0.0, Culture=neutral, PublicKeyToken=7cec85d7bea7798e]].IntroSort(Span`1 keys, Span`1 values, Int32 depthLimit, IComparer`1 comparer) in /_/src/libraries/System.Private.CoreLib/src/System/Collections/Generic/ArraySortHelper.cs:line 1429
   at System.Collections.Generic.ArraySortHelperForTComparer`3[[System.IntPtr, System.Private.CoreLib, Version=10.0.0.0, Cul

jkotas · 2025-10-01T00:37:26Z

Looks like a problem with #116109 (comment) . The special casing is needed to substitute null comparer with a default one.

jkotas · 2025-10-01T09:53:11Z

@EgorBot -intel

using BenchmarkDotNet.Attributes;

public class Bench
{
    int[] a = new int[10000];
    int[] b = new int[10000];
    MyComparer c = new MyComparer();

    [IterationSetup]
    public void IterationSetup() { for (int i = 0; i < a.Length; i++) a[i] = i; }

    [Benchmark]
    public void Sort() => Array.Sort(a, b, c);

    class MyComparer : IComparer<int>
    {
        public int Compare(int x, int y) => x-y;
    }   
}

tarekgh · 2025-10-02T23:11:25Z

Looks the benchmark numbers shared by @EgorBo not showing any regression. @jkotas what do you think?

jkotas · 2025-10-03T02:05:54Z

@EgorBot -intel -arm

using BenchmarkDotNet.Attributes;

public class Bench
{
    int[] a = new int[10000];
    int[] b = new int[10000];
    MyComparer<string> c = new MyComparer<string>();

    [IterationSetup]
    public void IterationSetup() { for (int i = 0; i < a.Length; i++) a[i] = i; }

    [Benchmark]
    public void Sort() => Array.Sort(a, b, c);

    struct MyComparer<T> : IComparer<int>
    {
        T state;
        long a,b,c,d,e,f;

        public int Compare(int x, int y) => x-y;
    }   
}

jkotas · 2025-10-03T03:23:00Z

LGTM (with the caveats listed in #116109 (comment)).

jkotas · 2025-11-20T06:22:28Z

@EgorBot -intel -arm -commit 1c955dd vs previous

using BenchmarkDotNet.Attributes;

public class Bench
{
    int[] a = new int[10000];
    int[] b = new int[10000];
    MyComparer<string> c = new MyComparer<string>();

    [IterationSetup]
    public void IterationSetup() { for (int i = 0; i < a.Length; i++) a[i] = i; }

    [Benchmark]
    public void Sort() => MemoryExtensions.Sort(a, b, c);

    struct MyComparer<T> : IComparer<int>
    {
        T state;
        long a,b,c,d,e,f;

        public int Compare(int x, int y) => x-y;
    }   
}

jkotas · 2025-11-20T06:25:53Z

@EgorBot -intel -arm -commit 1c955dd vs previous

using BenchmarkDotNet.Attributes;

public class Bench
{
    int[] a = new int[10000];
    int[] b = new int[10000];
    MyComparer c = new MyComparer();

    [IterationSetup]
    public void IterationSetup() { for (int i = 0; i < a.Length; i++) a[i] = i; }

    [Benchmark]
    public void Sort() => MemoryExtensions.Sort(a, b, c);

    class MyComparer : IComparer<int>
    {
        public int Compare(int x, int y) => x-y;
    }   
}

jkotas · 2025-11-28T22:21:57Z

Looks the EgorBot/runtime-utils#506 (comment) shared by @EgorBo not showing any regression. @jkotas what do you think?

@tarekgh The benchmark had mistake and it was not actually testing the code changed in this PR.

The updated results show up to ~3x regressions for comparers that are larger structs: EgorBot/runtime-utils#549 (comment) .

I think it is still fine to keep this change. Larger structs are frowned upon, so it should be ok to regress comparers that are larger structs and improve comparers that are small structs.

2A5F requested review from MichalStrehovsky, steveisok and vitek-karas as code owners May 29, 2025 18:39

github-actions bot added the needs-area-label An area label is needed to ensure this gets routed to the appropriate area owners label May 29, 2025

dotnet-policy-service bot added the community-contribution Indicates that the PR has been added by a community member label May 29, 2025

This was referenced May 29, 2025

slow macOS - "##[error]The job running on agent Azure Pipelines 9 ran longer than the maximum time of 60 minutes." dotnet/dnceng#1883

Open

The Operation will be canceled. The next steps may not contain expected logs. dotnet/dnceng#3008

Open

2A5F marked this pull request as draft May 30, 2025 11:03

2A5F marked this pull request as ready for review May 30, 2025 13:05

github-actions bot mentioned this pull request May 30, 2025

116109 MichalStrehovsky/rt-sz#127

Closed

EgorBot mentioned this pull request May 31, 2025

Benchmarks for #116109 (jkotas) EgorBot/runtime-utils#370

Open

jkotas added area-System.Memory tenet-performance Performance related issue and removed needs-area-label An area label is needed to ensure this gets routed to the appropriate area owners labels May 31, 2025

2A5F marked this pull request as draft May 31, 2025 09:29

2A5F force-pushed the avoid-sort-allocations branch from c46122b to 1b22a69 Compare June 2, 2025 15:07

2A5F requested a review from thaystg as a code owner September 25, 2025 04:26

dotnet-policy-service bot added the linkable-framework Issues associated with delivering a linker friendly framework label Sep 25, 2025

Extracted a separate region to prevent accidental modifications

25b1914

2A5F force-pushed the avoid-sort-allocations branch from 6c02cb3 to 25b1914 Compare September 25, 2025 04:32

jkotas reviewed Sep 25, 2025

View reviewed changes

src/libraries/System.Private.CoreLib/src/System/Collections/Generic/ArraySortHelper.cs Outdated Show resolved Hide resolved

jkotas reviewed Sep 25, 2025

View reviewed changes

src/libraries/System.Private.CoreLib/src/System/MemoryExtensions.cs Show resolved Hide resolved

jkotas reviewed Sep 25, 2025

View reviewed changes

src/libraries/System.Private.CoreLib/src/System/MemoryExtensions.cs Show resolved Hide resolved

2A5F added 2 commits September 25, 2025 13:12

ArraySortHelperForTComparer

6b40c5f

missing static

71d5295

This was referenced Sep 25, 2025

Checkout failure: "Git fetch failed with exit code 128" dotnet/arcade#9009

Open

Unable to checkout dotnet/dnceng#4115

Open

Merge branch 'main' into avoid-sort-allocations

e0e2029

restore special casing

5a858cd

EgorBot mentioned this pull request Oct 1, 2025

Benchmarks for #116109 (jkotas) EgorBot/runtime-utils#506

Open

EgorBot mentioned this pull request Oct 3, 2025

Benchmarks for #116109 (jkotas) EgorBot/runtime-utils#507

Open

jkotas approved these changes Oct 3, 2025

View reviewed changes

tarekgh merged commit 1c955dd into dotnet:main Oct 3, 2025
144 checks passed

dotnet-maestro bot mentioned this pull request Oct 14, 2025

[main] Source code updates from dotnet/runtime dotnet/dotnet#2885

Merged

github-actions bot locked and limited conversation to collaborators Nov 3, 2025

Avoid sort allocations #116109

Avoid sort allocations #116109

Uh oh!

Conversation

2A5F commented May 29, 2025

Uh oh!

jkotas commented May 31, 2025

Uh oh!

jkotas commented May 31, 2025

Uh oh!

2A5F commented May 31, 2025

Uh oh!

adamsitnik commented Jun 2, 2025

Uh oh!

2A5F commented Jun 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Test Code

Int32

BigStruct

IntClass

IntStruct

String

Uh oh!

2A5F commented Jun 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

andrewjsaid commented Jun 3, 2025

Uh oh!

2A5F commented Jun 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

andrewjsaid commented Jun 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

2A5F commented Jun 3, 2025

Uh oh!

andrewjsaid commented Jun 3, 2025

Uh oh!

2A5F commented Sep 25, 2025

Uh oh!

Uh oh!

Uh oh!

jkotas commented Sep 25, 2025

Uh oh!

Uh oh!

tarekgh commented Oct 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jkotas commented Oct 1, 2025

Uh oh!

jkotas commented Oct 1, 2025

Uh oh!

tarekgh commented Oct 2, 2025

Uh oh!

jkotas commented Oct 3, 2025

Uh oh!

jkotas commented Oct 3, 2025

Uh oh!

Uh oh!

jkotas commented Nov 20, 2025

Uh oh!

jkotas commented Nov 20, 2025

Uh oh!

jkotas commented Nov 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

2A5F commented Jun 2, 2025 •

edited

Loading

2A5F commented Jun 2, 2025 •

edited

Loading

2A5F commented Jun 3, 2025 •

edited

Loading

andrewjsaid commented Jun 3, 2025 •

edited

Loading

tarekgh commented Oct 1, 2025 •

edited

Loading

jkotas commented Nov 28, 2025 •

edited

Loading