CodSpeedHQ/codspeed-node - CodSpeed

codspeed-node

Blog Docs Changelog

Performance History

Latest Results

fixup! feat(tinybench): run walltime benchmarks in forks

cod-3960-implement-process-isolation-in-codspeed-node

5 hours ago

fixup! feat(tinybench): run walltime benchmarks in forks

cod-3960-implement-process-isolation-in-codspeed-node

6 hours ago

fixup! feat(vitest): run walltime benchmarks in forks

cod-3960-implement-process-isolation-in-codspeed-node

6 hours ago

feat(vitest): run walltime benchmarks in forks With Vitest 5, each walltime benchmark runs in fresh `vitest bench` processes restricted to its file. The fork count resolves as `CODSPEED_WALLTIME_FORKS`, the benchmark's or run's `forks` option, `codspeedPlugin({ forks })`, then 3. The test timeout is scaled by the env or plugin fork count, since every fork of a test runs within it. Vitest's own report shows zeroed statistics for forked benchmarks; the CodSpeed results hold the pooled rounds. Vitest 3/4 don't fork. Closes COD-3960

cod-3960-implement-process-isolation-in-codspeed-node

7 hours ago

fix(vitest-plugin): share the parent's global setup with walltime forks The parent's global setup stays up until it closes, after its forks, so a fork that ran it again collided with the resources it holds (fixed ports, databases, ...). Forks now skip the user's global setup and inherit the context it provided, handed down through a file in the environment.

cod-3960-implement-process-isolation-in-codspeed-node

8 hours ago

fix(vitest-plugin): share the parent's global setup with walltime forks The parent's global setup stays up until it closes, after its forks, so a fork that ran it again collided with the resources it holds (fixed ports, databases, ...). Forks now skip the user's global setup and inherit the context it provided, handed down through a file in the environment.

cod-3960-implement-process-isolation-in-codspeed-node

8 hours ago

fixup! feat(core): run walltime benchmarks in forks

cod-3960-implement-process-isolation-in-codspeed-node

11 hours ago

fixup! feat(vitest): run walltime benchmarks in forks Launch the forks from the main process once the run is over instead of inside the test. The parent worker ran the file's hooks before spawning, so a fork that needed the same exclusive resource (a port, a database) failed on it. The provider now only records the benchmarks on the test's meta, which Vitest ships to the main process with the result, and a reporter forks them from onTestRunEnd, after every file has torn down. Only the main process spawns now, so the forks re-run its own command: the recorded isolation command and the plugin forks env are gone, and the reporter resolves the fork count from the benchmark's option or the plugin's. Forks no longer run inside the test's timeout, so the fork-scaled test timeout and its tests are removed. An exclusive-port bench covers the overlap; its fn is async so the analysis path awaits each connect. Co-Authored-By: Claude <noreply@anthropic.com>

cod-3960-implement-process-isolation-in-codspeed-node

11 hours ago

Latest Branches

+66%

feat: run walltime benchmarks in forks#101

5 hours ago

18f0083

cod-3960-implement-process-isolation-in-codspeed-node

×3.3

perf(tinybench): batch fast walltime invocations into rounds#117

1 day ago

cfed9fe

cod-4104-batch-fast-tinybench-walltime-benchmarks-into-1-ms-rounds

+21%

docs(core): note that the CodSpeed app tracks the tested Node.js majors#116

2 days ago

e96240a

cod-3506-runtime-matrix-pointer

© 2026 CodSpeed Technology

Home Terms Privacy Docs