[CI] Improve benchmark outputs

[yzhangcs]

Summary by CodeRabbit

• Style

  • Updated benchmark comparison output to use emoji indicators (🔴 for regressions, 🟢 for speedups) instead of text markers.

• Chores

  • Enhanced benchmark result processing to display comprehensive results in PR comments, improving visibility of performance metrics across test configurations.

[coderabbitai]

Caution

Review failed

Pull request was closed or merged during review

Walkthrough

The pull request enhances benchmark result comparison infrastructure by replacing simple regression/speedup text markers with emoji indicators (🔴/🟢) and refactoring the GitHub Actions workflow to process detailed base and head result datasets, computing percent changes and speedup ratios in a comprehensive results table.

Changes

Cohort / File(s)	Summary
Benchmark workflow enhancement .github/workflows/reusable-ci-benchmarks.yml	Updated GitHub Actions script to process base_results and head_results instead of simplified arrays; constructs lookup maps keyed by operation parameters; generates full results table with percent change calculations, speedup ratios, and conditional red/green markers when thresholds are exceeded; handles new/removed result entries.
Benchmark comparison script scripts/run_benchmark_compare.py	Replaced text-based regression/speedup markers with emoji symbols (🔴 for regressions, 🟢 for speedups); removed shebang line.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• fla-org/flash-linear-attention#799: Modifies the same benchmark workflow and script files with overlapping CI benchmarking infrastructure changes.

Poem

🐰 Benchmarks now wear their colors bright,
Emoji markers—red and green—in sight!
Results bloom forth in tables grand,
Percent changes helping us understand! ✨

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The PR title '[CI] Improve benchmark outputs' accurately summarizes the main changes: enhancing the benchmark comparison workflow and script output formatting with better visual indicators and result handling.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch ci

---

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

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[gemini-code-assist]

Code Review

This pull request removes a misplaced shebang from the benchmark comparison script and replaces the text-based performance markers 'REGRESSION' and 'SPEEDUP' with emoji indicators for improved visual clarity. I have no feedback to provide.