[not-matthias]

TLDR: Batch fast Tinybench walltime calls into approximately 1 ms rounds to reduce timer and sample-processing overhead (stacked on #101).

• Estimate batch size from warmup, or calibrate after lifecycle setup when warmup is disabled. Later forks reuse the first fork's size.
• Normalize CodSpeed latency statistics per invocation; keep total measured time unscaled. Samples describe batch averages, not individual-call tails.
• Preserve sequential async calls. Per-invocation hooks and task concurrency stay unbatched; fallback, simulation, and Vitest measurement behavior are unchanged.

Local iterative Fibonacci(10), Node 24.22.1 / Tinybench 4.0.1, 1 s measurement + 100 ms warmup per process, one run per configuration:

Mode	Elapsed before → after	Samples before → after
Shared process	20.06 s → 1.11 s	28,006,582 → 891
Two forks	59.39 s → 10.33 s	58,596,190 → 1,760

Peak RSS of a measuring child in the two-fork run: 3,769 MiB → 71 MiB. Local smoke checks also cover Tinybench 6.2.2, lifecycle hooks, slow rounds, concurrency, fallback, and simulation.

Closes COD-4104.

[codspeed]

Merging this PR will regress 4 benchmarks

5 warnings, 1 note

⚠️ Walltime results may be unreliable

Some benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.

Learn more about Macro Runners

See affected benchmarks

⚠️ Unable to measure the execution of 8 benchmarks

These benchmarks measured no execution time, so their results are not comparable and count as unchanged. This usually happens when the compiler removes the code under test.

Prevent compiler optimizations

See affected benchmarks

⚠️ 14 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

⚠️ 7 benchmarks with a significant change ran on different hardware

Their base and head results were measured on machines with a different CPU, core count or memory, so the changes might not come from the code. Run base and head on the same runner type.

Learn more about Macro Runners

See affected benchmarks

⚠️ 7 benchmarks with a significant change ran in a different software environment

Their base and head results were measured with a different OS or toolchain, such as compiler, runtime or library versions, so the changes might come from this difference rather than from the code. Open an affected benchmark to see what changed.

See affected benchmarks

ℹ️ 16 benchmarks have no flamegraph

The flamegraph wasn't generated: most of the time was spent in runtime internals, which would leave nothing useful to analyze.

See affected benchmarks

⚡ 25 improved benchmarks
❌ 4 regressed benchmarks
✅ 240 untouched benchmarks
⏩ 57 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	CPU Simulation	wait 500ms	26.7 ms	38.4 ms	-30.59%
❌	CPU Simulation	wait 1sec	53.5 ms	72.9 ms	-26.68%
❌	Memory	one	1.5 KB	2 KB	-26.19%
❌	Walltime	fibo darwin	7.9 ms	8.8 ms	-10.39%
	Walltime	switch 2	30 ns	< 1 ns	×130
	Walltime	switch 2	30 ns	< 1 ns	×130
	Walltime	switch 2	30 ns	< 1 ns	×130
	Walltime	switch 1	30 ns	< 1 ns	×130
	Walltime	switch 2	30 ns	< 1 ns	×130
	Walltime	switch 1	30 ns	< 1 ns	×130
	Walltime	switch 1	30 ns	< 1 ns	×130
	Walltime	switch 1	30 ns	< 1 ns	×130
⚡	Walltime	test_iterative_fibo_10	30 ns	1.9 ns	×16
⚡	Walltime	test_iterative_fibo_10	30 ns	1.9 ns	×16
⚡	Walltime	test sync baz 10	30 ns	2.6 ns	×12
⚡	Walltime	test sync baz 10	30 ns	2.8 ns	×11
⚡	Walltime	test_iterative_fibo_100	60 ns	26.7 ns	×2.2
⚡	Walltime	test_iterative_fibo_100	60 ns	26.9 ns	×2.2
⚡	CPU Simulation	wait 500ms	54.8 ms	29.2 ms	+87.53%
⚡	Walltime	test sync baz 100	40 ns	22.5 ns	+77.99%
...	...	...	...	...	...

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-4104-batch-fast-tinybench-walltime-benchmarks-into-1-ms-rounds (38afc2a) with cod-3960-implement-process-isolation-in-codspeed-node (2e87449)}

Footnotes

1. 57 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩