Add necessary flags to dump optimized inlining info and fix walltime marker placement

[GuillaumeLagrange]

No description provided.

[codspeed-hq]

Merging this PR will degrade performance by 74.17%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 23 improved benchmarks
❌ 23 regressed benchmarks
✅ 180 untouched benchmarks
🆕 12 new benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Memory	switch 2	1 B	384 B	-99.74%
❌	Memory	switch 2	1 B	384 B	-99.74%
❌	Memory	test_iterative_fibo_100	1 B	264 B	-99.62%
❌	Memory	test_iterative_fibo_10	1 B	264 B	-99.62%
❌	Memory	test_recursive_cached_fibo_20	1 B	264 B	-99.62%
❌	Memory	test_recursive_cached_fibo_20	1 B	264 B	-99.62%
❌	Memory	test_recursive_cached_fibo_30	1 B	264 B	-99.62%
❌	Memory	test_iterative_fibo_100	1 B	264 B	-99.62%
❌	Memory	test sync baz 100	1 B	264 B	-99.62%
❌	Memory	test_recursive_cached_fibo_30	1 B	264 B	-99.62%
❌	Memory	switch 2	1 B	264 B	-99.62%
❌	Memory	test sync baz 100	1 B	264 B	-99.62%
❌	Memory	fibo 30	19 B	531 B	-96.42%
❌	Memory	test_recursive_fibo_20	19 B	264 B	-92.8%
❌	Memory	test_recursive_fibo_20	19 B	264 B	-92.8%
❌	Simulation	test_recursive_fibo_10	145.2 µs	282.2 µs	-48.56%
❌	Simulation	switch 2	11.2 µs	21.5 µs	-47.96%
❌	Memory	test sync baz 10	385 B	648 B	-40.59%
❌	Memory	test sync baz 10	385 B	648 B	-40.59%
❌	Simulation	one	402.7 µs	590.3 µs	-31.78%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing cod-2821-v8-maglevturbofan-inlining-loses-nested-inline-information (16bf967) with main (bcd7e64)¹}

Footnotes

1. No successful run was found on main (16bf967) during the generation of this report, so bcd7e64 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[greptile-apps]

Greptile Summary

This PR wires up --perf-prof and optional V8 log flags for walltime profiling with samply, and fixes the instrument window placement in both the tinybench and vitest plugins by driving startBenchmark/stopBenchmark from tinybench's suite-level setup/teardown hooks rather than wrapping Task.run directly.

• Core flags: getV8Flags gains a walltime case that emits --perf-prof (and log-file flags when CODSPEED_V8_LOG is set); linuxPerf.start/stop is skipped in walltime mode.
• tinybench-plugin refactor: task fn/fnOpts are now captured in a WeakMap at bench.add time (fixing tinybench v6 private-field incompatibility); the root frame is baked in at registration rather than at run time; installInstrumentHooks replaces the old wrapWithInstrumentHooks wrappers to bracket only the measured loop.
• vitest-plugin refactor: importTinybench now returns an InstrumentedBench subclass whose constructor installs per-bench hooks, and closeInstrumentWindow correctly emits markers before stopBenchmark to preserve the expected FIFO ordering.

Confidence Score: 5/5

The core logic changes are safe to merge; the marker-ordering issue in the tinybench walltime teardown was already identified in prior review rounds.

The V8 flag additions and linuxPerf gating are straightforward and low-risk. The tinybench-plugin refactor correctly replaces the broken private-field cast pattern with a WeakMap capture, and the vitest-plugin's closeInstrumentWindow already emits markers in the right order. The only unresolved finding in this pass is a missing Linear reference on a TODO comment in the workflow file.

packages/tinybench-plugin/src/walltime.ts — the marker-ordering issue (stopBenchmark before addMarker in the teardown hook) flagged in prior threads remains unaddressed in this diff.

Important Files Changed

Filename	Overview
.github/workflows/codspeed.yml	Adds samply profiler env var and pins an alpha runner version (v4.17.7-alpha.2) with a TODO that lacks a Linear issue reference
packages/tinybench-plugin/src/walltime.ts	Major refactor: drives instrumentation window via bench setup/teardown hooks; marker ordering (stopBenchmark before addMarker) has been flagged in prior review threads
packages/vitest-plugin/src/walltime/index.ts	Refactored to drive instrument window from per-bench setup/teardown hooks; closeInstrumentWindow correctly emits markers before stopBenchmark
packages/tinybench-plugin/src/analysis.ts	Switches to captured task data map instead of casting private task fields; adds optimizeFunctionSync warm-up pass to the sync path; no markers sent in analysis mode
packages/core/src/introspection.ts	Adds walltime case to getV8Flags switch: emits --perf-prof and optional V8 log-file flags when CODSPEED_V8_LOG is set; walltime case is missing break (flagged in prior thread)
packages/tinybench-plugin/src/benchOptions.ts	New helper to abstract v5 (bench.opts) vs v6 (bench directly) options layout, enabling in-place mutation of setup/teardown hooks
packages/tinybench-plugin/src/taskData.ts	New WeakMap-backed store capturing fn/fnOpts per task at bench.add time, replacing unsafe private-field casts that broke in tinybench v6
packages/tinybench-plugin/src/shared.ts	Removes wrapWithInstrumentHooks/wrapWithInstrumentHooksAsync and sendBenchmarkMarkers from base class; adds getTaskData helper
packages/tinybench-plugin/src/index.ts	Captures fn/fnOpts into taskDataMap at add time and wraps with wrapWithRootFrameSync in walltime mode to pre-bake the root frame before tinybench stores the function
.github/workflows/ci.yml	Excludes with-tinybench-v5 and with-tinybench-v6 from Node 18 matrix runs since tinybench v5+ requires Node >=20

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant TB as tinybench
    participant WR as WalltimeBenchRunner
    participant IH as InstrumentHooks (FIFO)

    Note over TB,IH: installInstrumentHooks() wires setup/teardown

    TB->>WR: opts.setup(task, "run")
    WR->>IH: startBenchmark()
    WR->>WR: "runStart = currentTimestamp()"
    TB->>TB: measured loop (task fn x N)
    TB->>WR: opts.teardown(task, "run")
    WR->>WR: "runEnd = currentTimestamp()"
    WR->>IH: stopBenchmark()
    WR->>IH: setExecutedBenchmark(pid, uri)
    WR->>IH: addMarker(BENCHMARK_START, runStart)
    WR->>IH: addMarker(BENCHMARK_END, runEnd)

    Note over WR,IH: tinybench plugin - markers land after StopBenchmark
    Note over WR,IH: vitest plugin closeInstrumentWindow - markers before StopBenchmark

Loading

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
sequenceDiagram
    participant TB as tinybench
    participant WR as WalltimeBenchRunner
    participant IH as InstrumentHooks (FIFO)

    Note over TB,IH: installInstrumentHooks() wires setup/teardown

    TB->>WR: opts.setup(task, "run")
    WR->>IH: startBenchmark()
    WR->>WR: "runStart = currentTimestamp()"
    TB->>TB: measured loop (task fn x N)
    TB->>WR: opts.teardown(task, "run")
    WR->>WR: "runEnd = currentTimestamp()"
    WR->>IH: stopBenchmark()
    WR->>IH: setExecutedBenchmark(pid, uri)
    WR->>IH: addMarker(BENCHMARK_START, runStart)
    WR->>IH: addMarker(BENCHMARK_END, runEnd)

    Note over WR,IH: tinybench plugin - markers land after StopBenchmark
    Note over WR,IH: vitest plugin closeInstrumentWindow - markers before StopBenchmark

Loading

_{Reviews (10): Last reviewed commit: "feat: to fixup tinybench marker clean up" | Re-trigger Greptile}