feat: add MacOS support

[not-matthias]

No description provided.

[codspeed-hq]

Merging this PR will degrade performance by 93.19%

⚠️ 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

⚡ 5 improved benchmarks
❌ 19 regressed benchmarks
✅ 201 untouched benchmarks
🆕 1 new benchmark

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	iterative fibo 15	312 ns	369,169 ns	-99.92%
❌	WallTime	iterative fibo 20	360 ns	367,274 ns	-99.9%
❌	WallTime	short body	1.7 µs	377 µs	-99.55%
❌	WallTime	short body	1.7 µs	379.6 µs	-99.55%
❌	WallTime	short body	1.7 µs	373.4 µs	-99.54%
❌	WallTime	short body	1.8 µs	373.7 µs	-99.53%
❌	WallTime	short body	1.8 µs	373.8 µs	-99.53%
❌	WallTime	short body	1.8 µs	376.5 µs	-99.52%
❌	WallTime	fibo 10	2.1 µs	372.6 µs	-99.43%
❌	WallTime	fibo 15	21.5 µs	392.1 µs	-94.51%
❌	WallTime	recursive fibo 15	22.2 µs	396.5 µs	-94.41%
❌	WallTime	end	35.5 µs	432 µs	-91.79%
❌	WallTime	long body	214.3 µs	604.3 µs	-64.53%
❌	WallTime	long body	211.7 µs	596.8 µs	-64.52%
❌	WallTime	long body	217.2 µs	601.9 µs	-63.91%
❌	WallTime	long body	217.3 µs	601.5 µs	-63.87%
❌	WallTime	recursive fibo 20	239.5 µs	613.1 µs	-60.94%
❌	Simulation	test_iterative_fibo_10	299.9 µs	547.2 µs	-45.2%
❌	WallTime	wait 1ms	1 ms	1.4 ms	-28.18%
⚡	Memory	wait 1ms	16 B	10 B	+60%
...	...	...	...	...	...

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-2720-bump-instrument-hooks-in-codspeed-node-to-support-macos (ad2b2e7) with main (4dae798)}

[greptile-apps]

Greptile Summary

This PR adds macOS walltime support to the vitest and tinybench plugins by introducing per-iteration BENCHMARK_START/BENCHMARK_END FIFO markers in WalltimeRunner and a single per-run marker pair in the tinybench BaseBenchRunner, backed by a submodule bump to expose currentTimestamp and addMarker. A new codspeed-walltime-macos CI job is added that targets the sip-resign-exec-redirect runner, and a dedicated macOS-only bench file guards the darwin benchmark with describe.skipIf.

• vitest-plugin/src/walltime/index.ts: __codspeed_root_frame__ now calls finishRound (2 FIFO round-trips) per iteration — this is new behaviour on the existing Linux codspeed-walltime path, not just macOS.
• tinybench-plugin/src/shared.ts: sendRunMarkers emits one marker pair after all tasks complete (correctly walltime-gated) to avoid the per-iteration FIFO overflow documented in its inline comment.
• ci.yml: Removes the old walltime-macos-test CI job; the macOS benchmark is now owned by codspeed.yml.

Confidence Score: 4/5

Safe to merge for the macOS-only path; the existing Linux walltime benchmarks now silently emit per-iteration FIFO markers that were not there before, matching a pattern the tinybench plugin explicitly removed to prevent CI timeouts.

The finishRound change in WalltimeRunner adds two blocking FIFO round-trips per tinybench iteration on every walltime run — including the four existing bench files on the Linux codspeed-walltime job. The tinybench plugin's own inline comment documents that this identical per-iteration pattern previously overwhelmed the runner and timed out CI, and that plugin was specifically refactored to avoid it. The macOS job uses a purpose-built runner branch, but the Linux job uses the existing runner with no changes and has no guard to skip the new markers.

packages/vitest-plugin/src/walltime/index.ts — the finishRound call inside codspeed_root_frame affects all walltime benchmark runs, not only the new macOS benchmark.

Important Files Changed

Filename	Overview
packages/vitest-plugin/src/walltime/index.ts	Adds per-iteration FIFO markers (BENCHMARK_START/END) in codspeed_root_frame; this is new behaviour on the Linux walltime path and contradicts the design rationale documented in the tinybench plugin.
packages/tinybench-plugin/src/shared.ts	Adds sendRunMarkers emitting a single BENCHMARK_START/END pair bracketing the whole run; correctly guards on walltime mode and documents the per-iteration marker problem.
packages/tinybench-plugin/src/walltime.ts	Switches warmup call from task.warmup() to task.warmupSync() for the sync path; cosmetic blank-line addition.
packages/tinybench-plugin/tests/index.integ.test.ts	Adds currentTimestamp and addMarker mocks required by the new sendRunMarkers code path.
packages/vitest-plugin/benches/macos.bench.ts	New macOS-only benchmark file; uses describe.skipIf(!isMacOS) to guard the darwin-only fibo bench.
.github/workflows/codspeed.yml	Adds codspeed-walltime-macos job; still contains the diagnostic resign+exec step and a mutable branch:sip-resign-exec-redirect runner-version (both already flagged in previous threads).
.github/workflows/ci.yml	Removes the old walltime-macos-test job; no new logic introduced.
packages/core/src/native_core/instruments/hooks	Submodule bumped to b9ddb5b; new commit adds currentTimestamp and addMarker FIFO support required by this PR.

Sequence Diagram

sequenceDiagram
    participant TB as Tinybench Task.run()
    participant CF as __codspeed_root_frame__
    participant FN as benchmark fn()
    participant FIFO as Runner FIFO

    TB->>FIFO: startBenchmark() [ACK]
    loop N iterations
        TB->>CF: call fn
        CF->>CF: "roundStart = currentTimestamp()"
        CF->>FN: fn()
        FN-->>CF: result
        CF->>FIFO: addMarker(BENCHMARK_START, roundStart) [ACK]
        CF->>FIFO: addMarker(BENCHMARK_END, roundEnd) [ACK]
        CF-->>TB: result
    end
    TB->>FIFO: stopBenchmark() [ACK]
    TB->>FIFO: setExecutedBenchmark(pid, uri) [ACK]

Loading

_{Reviews (9): Last reviewed commit: "fix(ci): only run the macOS-only bench o..." | Re-trigger Greptile}

[greptile-apps]

The multi-clause thenable check should be extracted into a named const per the project's rule that non-obvious or multi-clause boolean conditions must be named with an is*/has* prefix.

Suggested change

	if (
	result !== null &&
	typeof result === "object" &&
	typeof (result as PromiseLike<unknown>).then === "function"
	) {
	const isThenable =
	result !== null &&
	typeof result === "object" &&
	typeof (result as PromiseLike<unknown>).then === "function";
	if (isThenable) {

Rule Used: Extract any non-obvious or multi-clause boolean co... (source)

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

[greptile-apps]

Diagnostic step left in production workflow

The "Diagnose resign+exec on this runner" step copies /bin/sh, codesigns it, and re-executes it — this is clearly debugging scaffolding for the SIP/resign investigation. Leaving it in the permanent workflow adds noise to every future CI run and exposes internal implementation details. It should be removed once the resign-exec behaviour is confirmed working.

[greptile-apps]

Mutable branch reference for runner-version

runner-version: branch:sip-resign-exec-redirect points to a live branch that can be force-pushed or rebased at any time. If the branch tip changes between two benchmark runs, the profiling behaviour changes silently, making measurement results non-comparable across commits. Consider pinning to a specific commit SHA or tag once the feature branch is stable, e.g. runner-version: sha:<commit>.