run-benchmarks

name	run-benchmarks
description	Run Rask render/runtime benchmarks before and after a framework change and report the Allocated delta. Use whenever you modify render hot-path or live-runtime code in src/Rask.Core or src/Rask.Server (HtmlSerializer, Element/Component render, Live/* diff codec, WS/WASM dispatch, payload build). Required evidence for any hotpath PR.

run-benchmarks

BenchmarkDotNet, Release only, hand-run (not part of dotnet test). Project: benchmarks/Rask.Benchmarks/Rask.Benchmarks.csproj. All benches use [MemoryDiagnoser].

1. Pick the bench class matching the change

Area changed	Bench class (--filter "*Name*")
End-to-end HTML render + WS payload	RenderRoundTripBenchmarks, LiveRenderRoundTripBenchmarks
Payload build / body inject/extract	LivePayloadUtf8Benchmarks
Diff DOM-walk	FrameDifferBenchmarks
Attribute encoding	AttributeEncodingBenchmarks
WS / WASM dispatch	WsDispatchBenchmarks, WasmDispatchBenchmarks
Asset / download	AssetLoadingBenchmarks, DownloadPayloadBenchmarks

2. Baseline (pre-change) then post-change

# capture baseline on the unchanged tree
git stash                      # or check out the parent commit
dotnet run -c Release --project benchmarks/Rask.Benchmarks/Rask.Benchmarks.csproj --filter "*RenderRoundTrip*"
# results land in benchmarks/Rask.Benchmarks/BenchmarkDotNet.Artifacts/results/*-report-github.md
git stash pop                  # restore the change, re-run the same filter
dotnet run -c Release --project benchmarks/Rask.Benchmarks/Rask.Benchmarks.csproj --filter "*RenderRoundTrip*"

3. Compare + report

Read the Allocated (and Mean) columns from the two BenchmarkDotNet.Artifacts/results/*-report-github.md runs. Trust Allocated even on InvocationCount=1 benches. Quote the delta (e.g. "Allocated 1.84 KB → 0.91 KB, −51%") in the PR body. Custom reports: append bundle-size, payload-bytes, session-footprint (per-live-session retained memory + sessions-per-GiB across a page-size sweep) or session-churn (soak + create/dispose churn) instead of --filter.