Migrate to .NET 10 from .NET 8 by Carti-it · Pull Request #1333 · MetacoSA/NBitcoin

Carti-it · 2026-09-07T08:10:19Z

One thing that I cannot really explain is this:

master (net8.0)

cd NBitcoin.Bench
dotnet run -f net8.0 -c Release -- --runtimes net8.0 --filter *HexBench*

Method	hexString	Mean	Error	StdDev	Gen0	Allocated
DecodeData	?	25.11 ns	0.434 ns	0.385 ns	0.0089	56 B
DecodeDataLongSpan	?	4,030.49 ns	79.167 ns	102.940 ns	1.6289	10264 B
EncodeData	?	64.81 ns	1.324 ns	2.583 ns	0.0485	304 B
DecodeDataSpan	00000(...)fffff [64]	32.56 ns	0.679 ns	1.171 ns	0.0089	56 B
DecodeDataSpan	0000(...)ffff [128]	47.32 ns	0.967 ns	1.588 ns	0.0140	88 B
DecodeDataSpan	0000(...)ffff [192]	64.95 ns	1.325 ns	3.780 ns	0.0191	120 B

PR (net10.0)

cd NBitcoin.Bench
dotnet run -f net10.0 -c Release -- --runtimes net10.0 --filter *HexBench*

Method	hexString	Mean	Error	StdDev	Gen0	Allocated
DecodeData	?	40.50 ns	0.704 ns	0.659 ns	0.0089	56 B
DecodeDataLongSpan	?	9,580.75 ns	166.571 ns	155.811 ns	-	-
EncodeData	?	17.18 ns	0.276 ns	0.258 ns	0.0242	152 B
DecodeDataSpan	00000(...)fffff [64]	35.94 ns	0.402 ns	0.356 ns	-	-
DecodeDataSpan	0000(...)ffff [128]	64.46 ns	1.107 ns	0.981 ns	-	-
DecodeDataSpan	0000(...)ffff [192]	99.97 ns	1.786 ns	1.671 ns	-	-

(This was measured with the latest BenchmarkDotNet version (v0.15.8) and I did the same for .NET 8. So that's not the reason.)

Decoding HEX on .NET 10 seems to be slower than on .NET 8. I think that all other perf improvements of .NET 10 outweighs this one but still I don't really know what the issue might be. I kept playing with it for a while but I don't know.

Perhaps

it's a really a genuine regression in .NET 10,
or it needs to be reproduced on a different machine than mine
or perhaps something like dotnet run -f net8.0 -c Release -- --runtimes net10.0,net8.0 --filter *HexBench* would be more fair comparison.

I don't know.

Carti-it · 2026-09-07T08:11:16Z

 			using (var fs = File.Open(filePath, FileMode.Open, FileAccess.Read))
 			{
 				data = new byte[fs.Length - 32];
-				fs.Read(data, 0, data.Length);


This reports an analyzer warning because it can return after a partial read.

Carti-it · 2026-09-07T08:11:57Z

 	</PropertyGroup>
 	<PropertyGroup>
-    <TargetFrameworks>net472;netstandard2.1;netstandard2.0;net8.0</TargetFrameworks>
+    <TargetFrameworks>net10.0;net472;netstandard2.1;netstandard2.0</TargetFrameworks>


net10.0 TFM was moved everywhere to be the first entry in <TargetFrameworsk> to unify it in code.

Carti-it · 2026-09-07T08:12:34Z

@@ -1,4 +1,4 @@
-#if !NET6_0_OR_GREATER && !NO_TUPLE
+#if !NET10_0_OR_GREATER && !NO_TUPLE


Easier to find in the future

Carti-it · 2026-09-07T08:12:54Z

@@ -6,47 +6,47 @@ env:
  DOTNET_CLI_TELEMETRY_OPTOUT: 'true'


Done based on fa4e677

Carti-it · 2026-09-07T08:13:09Z

  <PropertyGroup>
    <OutputType>Exe</OutputType>
-    <TargetFrameworks>net8.0</TargetFrameworks>
+    <TargetFrameworks>net10.0</TargetFrameworks>


Done based on dc5118a

Carti-it · 2026-09-07T08:13:39Z

  <ItemGroup>
-    <PackageReference Include="BenchmarkDotNet" Version="0.14.0" />
-    <PackageReference Include="BenchmarkDotNet.Diagnostics.Windows" Version="0.14.0" />
+    <PackageReference Include="BenchmarkDotNet" Version="0.15.8" />


Required for .NET 10 support

Carti-it · 2026-09-07T08:13:55Z

 		/// </summary>
 		public ExtPubKey(ReadOnlySpan<byte> bytes)
 		{
-			if (bytes == null)


Warning: span cannot be null

Carti-it · 2026-09-07T08:14:13Z

 #endif

+#if NET10_0_OR_GREATER
+		public override string EncodeData(ReadOnlySpan<byte> data)


Added to avoid byte[] allocations.

…ffers to 4.6.1 from 4.5.0

Carti-it · 2026-09-07T08:37:24Z

 		<PackageReference Include="Microsoft.CSharp" Version="4.7.0" />
-		<PackageReference Include="xunit" Version="2.9.2" />
-		<PackageReference Include="xunit.runner.visualstudio" Version="2.4.1">
+		<PackageReference Include="xunit" Version="2.9.3" />


xunit package is deprecated.

WalletWasabi/WalletWasabi#14864 - upgrade to xUnit v3 is not that hard

NicolasDorier · 2026-09-07T09:09:04Z

@Carti-it I find it weird that 8.0 was faster.
Did you try just bumping 10.0 without the call to Convert.ToHexStringLower? Might it be the code path isn't used?

Carti-it · 2026-09-07T09:18:51Z

@Carti-it I find it weird that 8.0 was faster.

Yes. It's unexpected. I would expect better times, or at worst the same times. Certainly not worse.

Did you try just bumping 10.0 without the call to Convert.ToHexStringLower?

Yes. It was also slower. About the same as now.

Might it be the code path isn't used?

I run tests with the debugger on and the .NET 10 code paths get hit. So I don't think this is the issue.

Carti-it · 2026-09-07T09:19:51Z

@NicolasDorier Perhaps you can try to measure it on your machine.

Carti-it · 2026-09-07T13:37:37Z

dotnet/runtime#133350

NicolasDorier · 2026-09-08T03:36:02Z

that was fast, even pinpointed on dotnet/runtime wow

Carti-it added 6 commits September 7, 2026 10:08

Migrate to .NET 10

755e309

Upgrade preprocessor directives

39e249a

ExtPubKey: Span cannot be null

42dfc15

AddressManager: Make sure to read all bytes

e49a2d7

HexEncoder: Optimize EncodeData to avoid byte[] allocations

20139ba

Typo

4fb02ce

Carti-it commented Sep 7, 2026

View reviewed changes

Carti-it mentioned this pull request Sep 7, 2026

Optimize Encoders.Hex.EncodeData(ReadOnlySpan<byte> data) #1332

Closed

Carti-it added 2 commits September 7, 2026 10:19

NBitcoin: Upgrade Newtonsoft.Json to 13.0.4 from 13.0.1 and System.Bu…

2c92e92

…ffers to 4.6.1 from 4.5.0

NBitcoin.Tests: Upgrade packages

d392eb7

Carti-it commented Sep 7, 2026

View reviewed changes

NicolasDorier merged commit c7dcbd3 into MetacoSA:master Sep 8, 2026
6 checks passed

Carti-it deleted the net10 branch September 8, 2026 06:03

Carti-it mentioned this pull request Sep 21, 2026

RpcClient: Use NBitcoin's "GetBlock with verbosity 3" implementation WalletWasabi/WalletWasabi#14949

Merged

		@@ -1,4 +1,4 @@
		#if !NET6_0_OR_GREATER && !NO_TUPLE
		#if !NET10_0_OR_GREATER && !NO_TUPLE

Conversation

Carti-it commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

master (net8.0)

PR (net10.0)

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

NicolasDorier commented Sep 7, 2026

Uh oh!

Carti-it commented Sep 7, 2026

Uh oh!

Carti-it commented Sep 7, 2026

Uh oh!

Carti-it commented Sep 7, 2026

Uh oh!

NicolasDorier commented Sep 8, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Carti-it commented Sep 7, 2026 •

edited

Loading