Parent Issue

Part of #315 (TCP receive std::span callback migration)

Summary

Add a focused benchmark that quantifies the overhead difference between:

• span-based receive dispatch (no per-iteration allocation)
• legacy vector-based receive dispatch (per-iteration allocation + copy)

This provides a concrete baseline for the TCP receive hot-path migration and helps detect regressions in CI.

Background

After #316, tcp_socket can deliver received bytes as std::span<const uint8_t> without allocating/copying into a temporary vector. We should add a micro benchmark to validate the expected performance gap and keep it visible over time.

Proposed Benchmarks

Add a new benchmark file (or extend an existing one):

• benchmarks/tcp_receive_bench.cpp (preferred) or extend benchmarks/message_throughput_bench.cpp

Benchmarks:

1. BM_TcpReceive_Dispatch_Span

   • Create a fixed std::array<uint8_t, N>
   • Create a std::span<const uint8_t> view
   • Invoke a trivial callback that touches bytes (to prevent dead-code elimination)
   • No vector construction

2. BM_TcpReceive_Dispatch_VectorFallback

   • Same fixed input
   • Construct std::vector<uint8_t> each iteration (simulating legacy adapter path)
   • Invoke the same callback signature

Optional:

• Run with --benchmark_min_time=5.0 for stable results in CI.

Files to Update

• benchmarks/CMakeLists.txt
• Add benchmarks/tcp_receive_bench.cpp (or modify existing benchmark sources)

Acceptance Criteria

• Benchmarks compile under -DNETWORK_BUILD_BENCHMARKS=ON
• Benchmark binary includes the new TCP receive dispatch benchmarks
• CI benchmark workflow runs successfully and publishes results
• Results clearly show span path outperforming vector fallback for small payloads

Test Plan

Local:

cmake -B build -S . -DNETWORK_BUILD_BENCHMARKS=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build
./build/benchmarks/network_benchmarks --benchmark_filter=TcpReceive --benchmark_min_time=3.0

Notes

This benchmark intentionally isolates dispatch/copy costs; it does not attempt to benchmark kernel/network I/O.