docs: explain benchmark output fields#888
Conversation
Document mean ± σ, min/max range, User/System CPU times, and relative speed comparison in the man page and README. Co-authored-by: Cursor <cursoragent@cursor.com>
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 62883fbb50
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
|
||
| When comparing multiple commands, hyperfine prints a relative speed summary at the | ||
| end, for example `2.00 ± 0.05 times faster than …`. The value is the ratio of mean | ||
| runtimes; the `±` suffix shows the standard deviation of that ratio across runs. |
There was a problem hiding this comment.
The new README text says the relative ± value is “the standard deviation of that ratio across runs,” but compute_relative_speeds derives it via uncertainty propagation from each command’s mean/stddev (ratio * sqrt((σ_a/μ_a)^2 + (σ_b/μ_b)^2) with independence assumed) rather than by computing per-run ratios. This wording can mislead users about what the confidence term represents and should be corrected to match the implemented calculation.
Useful? React with 👍 / 👎.
1 participant
Summary
doc/hyperfine.1describing mean ± σ, min/max range, User/System CPU times, and relative speed comparisonFixes #443
Test plan