🤖 Update Benchmark Results

[thomhurst]

Automated Benchmark Update

This PR updates the benchmark documentation with the latest results from the Speed Comparison workflow.

Benchmarks Produced

Individual benchmark artifacts are available for download:

• benchmark-DataDrivenTests
• benchmark-AsyncTests
• benchmark-ScaleTests
• benchmark-MatrixTests
• benchmark-MassiveParallelTests
• benchmark-SetupTeardownTests
• benchmark-BuildTime
• benchmark-summary (aggregated overview)

Runtime Benchmarks

• AsyncTests
• DataDrivenTests
• MassiveParallelTests
• MatrixTests
• ScaleTests
• SetupTeardownTests

Build Benchmarks

• BuildTime
• 2026-09-29
• Ubuntu Latest • .NET SDK 10.0.401

Changes

• Updated benchmark data in docs/static/benchmarks/latest.json
• Updated historical trends in docs/static/benchmarks/historical.json
• Regenerated benchmark documentation in docs/docs/benchmarks/index.md
• Updated benchmark summary in docs/static/benchmarks/summary.json

Workflow Run

• Run ID: 36608749549
• Triggered: workflow_dispatch
• Date:

---

🤖 This PR was automatically created and will be merged automatically once CI checks pass.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-29T18:20:06.593572Z	5b9ba94	PR opened

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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

Review

Automated benchmark refresh, docs/data only (README table, benchmark markdown, static JSON). No source or engine code is touched, so there are no correctness, AOT or API concerns.

• The README table and versions (TUnit 1.72.0) are consistent with the run date (2026-09-29).
• Relative ordering is unchanged: TUnit (AOT) is fastest and TUnit beats the other frameworks in every scenario.
• Absolute times for every framework moved together, up in most scenarios and down in the scale suite. That looks like CI runner noise rather than a regression.
• I did not validate the JSON files programmatically, since a shell loop was blocked in this environment. I only spot-checked the diff.

LGTM.

[greptile-apps]

Confidence Score: 4/5

[Low risk] Benchmark results and documentation updated with new measurements.

The PR appears safe to merge, with a non-blocking history-retention issue worth correcting.

Findings

1.  Duplicate history date ▶

Summary

The PR refreshes benchmark measurements and TUnit version labels across the README, benchmark pages, and published JSON data.

• The generated history adds a duplicate same-day entry, reducing distinct-date coverage under its retention cap.

_{Reviews (1) · Last reviewed commit: "chore: update benchmark results"}

[greptile-apps]

Duplicate history date This update adds a second entry for 2026-09-29. The history keeps only the latest 90 entries, so the duplicate takes a slot and pushes out an older distinct date. Repeated same-day runs will further reduce the dates available for future trend analysis; the history generator should replace or deduplicate same-day entries.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 5b9ba947b2

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".

[chatgpt-codex-connector]

Avoid appending duplicate same-day history records

When the benchmark workflow runs more than once on the same UTC date, as happened here after f197760, this adds a second indistinguishable 2026-09-29 entry. Because the producer subsequently truncates the array with slice(-90), the duplicate also evicts a genuine older data point, reducing historical coverage without recording a new time point; replace or deduplicate the current day's entry before truncation.

Useful? React with 👍 / 👎.