Table of Contents

Benchmarks

NexNet is benchmarked using BenchmarkDotNet to measure invocation overhead, throughput, and memory allocation.

Environment

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26100.7462/24H2/2024Update/HudsonValley)
Intel Core i7-10700 CPU 2.90GHz, 1 CPU, 16 logical and 8 physical cores
.NET SDK 9.0.305
  [Host]     : .NET 9.0.9 (9.0.9, 9.0.925.41916), X64 RyuJIT x86-64-v3
  Job-FJSVHN : .NET 9.0.9 (9.0.9, 9.0.925.41916), X64 RyuJIT x86-64-v3

Platform=X64  Runtime=.NET 9.0

Results

Method Mean Error StdDev Op/s Allocated
InvocationNoArgument 36.7 us 0.33 us 0.31 us 27,241.7 595 B
InvocationUnmanagedArgument 37.4 us 0.52 us 0.48 us 26,769.8 649 B
InvocationUnmanagedMultipleArguments 37.3 us 0.28 us 0.25 us 26,800.8 700 B
InvocationNoArgumentWithResult 36.9 us 0.35 us 0.32 us 27,095.2 633 B
InvocationWithDuplexPipe_Upload 51.8 us 0.60 us 0.51 us 19,295.4 14,951 B

Key Observations

  • Basic invocations achieve approximately 27,000 operations per second
  • Adding arguments increases allocation modestly (595 B → 700 B) with negligible latency impact
  • Return values add minimal overhead compared to void invocations
  • Duplex pipe operations are slower due to the additional pipe setup and streaming infrastructure, but still achieve ~19,000 ops/sec

Running Benchmarks

The benchmark project is located at src/NexNetBenchmarks/. Run with:

dotnet run -c Release --project src/NexNetBenchmarks/NexNetBenchmarks.csproj

See Also

Edit this page