[hmdsefi]

Adds TestPointReleaseCost to runner/toolexec/sched, the measurement XTE §9 asks for. It runs only with SCHED_COST=1 and skips otherwise, so CI does not run it.

The test times one scheduling point and its release, with the driver's synctest.Wait, at stack depths 0 and 40. On an Apple M4 Max with go1.27.3 it logs about 2.7 µs and 11.7 µs, under the spec's targets of 3.0 µs and 16 µs (not gated).

The same run measured the slowdown of the exact backend against plain builds, 100 seeds, median of 3: 1.21× for the root package and shims/netsim, 1.61× for the key-value corpus, 1.01× for the HTTP corpus, and 3.12× for the HTTP corpus with net/http rewritten, all under the 5× limit. The backend's code size is 2,718 lines of the 3,000 budget.

Closes #348

[coderabbitai]

Important

• Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

• Source mode: Auto
• Repository: hmdsefi/faultline
• Branch: plan/2.5-task-12
• File: .coderabbit.yaml
• Commit: 8fc1f80
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Advanced
• Run ID: e343f967-355d-424d-a84e-78c525fcbd01

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn