perf: cache native dispatch invokers

[Jim8y]

Description

This change removes the reflection-based hot path from native contract dispatch on master-n3.

InteropDescriptor and ContractMethodMetadata now build cached compiled invokers once, and the runtime dispatch loops in ApplicationEngine.OnSysCall(...) and NativeContract.Invoke(...) reuse pooled object?[] buffers instead of allocating fresh arrays and lists on every invocation.

The optimization keeps the previous behavior intact, including TargetInvocationException wrapping semantics, and adds unit coverage for the new invoker layer.

Change Log

• Add delegate InteropInvoker
• Add delegate NativeMethodInvoker
• Update class InteropDescriptor
• Update class ContractMethodMetadata
• Update method ApplicationEngine.OnSysCall(...)
• Update method NativeContract.Invoke(...)
• Add file tests/Neo.UnitTests/SmartContract/UT_InteropDescriptor.cs
• Update file tests/Neo.UnitTests/SmartContract/Native/UT_ContractMethodAttribute.cs

Fixes # (issue) - no GitHub issue is linked for this profiling finding.

Type of change

• Optimization (the change is only an optimization)
• Style (the change is only a code style for better maintenance or standard purpose)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

How Has This Been Tested?

• Unit Testing
• Run Application
• Local Computer Tests
• No Testing

Verification commands:

• dotnet test tests/Neo.UnitTests/Neo.UnitTests.csproj --filter UT_InteropDescriptor --logger "console;verbosity=minimal"
• dotnet test tests/Neo.UnitTests/Neo.UnitTests.csproj --filter UT_ContractMethodAttribute --logger "console;verbosity=minimal"
• dotnet test tests/Neo.UnitTests/Neo.UnitTests.csproj --logger "console;verbosity=minimal"

Benchmark results from an external harness that executes GAS.totalSupply() through the real native dispatch path (200 executions x 1000 calls = 200000 native dispatches):

• base (origin/master-n3): elapsed_ms=2457.30, allocated_bytes=1001359680
• fixed (fix-native-dispatch-delegates): elapsed_ms=2240.74, allocated_bytes=964599232

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• Any dependent changes have been merged and published in downstream modules

[codecov]

Codecov Report

❌ Patch coverage is 99.02913% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 78.89%. Comparing base (f7bc967) to head (6a31470).

Files with missing lines	Patch %	Lines
src/Neo/SmartContract/InteropDescriptor.cs	96.15%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@              Coverage Diff              @@
##           master-n3    #4518      +/-   ##
=============================================
+ Coverage      78.76%   78.89%   +0.13%     
=============================================
  Files            236      236              
  Lines          16283    16365      +82     
  Branches        2361     2373      +12     
=============================================
+ Hits           12825    12911      +86     
+ Misses          2708     2700       -8     
- Partials         750      754       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ajara87]

Wouldn't it be better to move this into the finally block so that if a failure occurs the parameters don't remain on the stack?

[Jim8y]

but this can not be "fixed" as its the exsiting logic, this pr shall not change any exsiting logic

[shargon]

We need benchmarks, maybe the expression is slower

[Jim8y]

Environment

• BenchmarkDotNet v0.15.8
• .NET SDK 10.0.100
• Runtime: .NET 10.0.0
• OS: Linux Ubuntu 24.04.4 LTS
• CPU: Intel Core Ultra 9 285K 3.69GHz
• Job: ShortRun (IterationCount=3, LaunchCount=1, WarmupCount=3)

Summary

Method	master-n3 Mean	PR Mean	Mean Delta	master-n3 Alloc	PR Alloc	Alloc Delta
Syscall_NoArgs	71.32 ns	65.82 ns	-5.50 ns (-7.71%)	118 B	94 B
-24 B (-20.34%)
Syscall_OneArg	111.17 ns	111.92 ns	+0.75 ns (+0.67%)	182 B	150 B
-32 B (-17.58%)
Native_NoArgs	130.27 ns	125.26 ns	-5.01 ns (-3.85%)	128 B	96 B
-32 B (-25.00%)
Native_OneArg	193.77 ns	184.15 ns	-9.62 ns (-4.96%)	272 B	152 B
-120 B (-44.12%)

[shargon]

Syscall_OneArg how can be less percentage if it's possitive?

[Jim8y]

Syscall_OneArg how can be less percentage if it's possitive?

its not, but i think 0.xx% should be considered a reasonable margin of error?

[erikzhang]

Have we started optimizing operations at the nanosecond level?

[Jim8y]

Have we started optimizing operations at the nanosecond level?

I am basically still focusing on optimizing the performance of resync the whole ledger, which will cost hours even days, native contract invoking is one of the hotpath of that. Another hotpath is persistance.

[Wi1l-B0t]

Merge these two if?

[shargon]

Agree