OpenParable
Skip to content
← Back to skills

Render Bench

Testing & Quality
render-bench

Render Bench is an Agent Skill for testing & quality published in the erweixin/RaTeX repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Run RaTeX rendering performance benchmarks (100-formula PNG/SVG/PDF, cold-vs-hot font cache timing, phase breakdown). Use when profiling render performance or comparing before/after optimization changes.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/erweixin/RaTeX/blob/HEAD/skills/render-bench/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/render-bench, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Render performance benchmarks

Cross-tool layout (Cursor / Claude Code / Codex)

  • Canonical copy: repository root skills/render-bench/SKILL.md (this file).
  • Tool entry points (symlinks to this file in-repo):
    • Cursor: .cursor/skills/render-bench/SKILL.md
    • Claude Code: .claude/skills/render-bench/SKILL.md
    • Codex: .agents/skills/render-bench/SKILL.md

When to use

  • After changes to rendering paths (especially renderer.rs, fonts.rs, standalone.rs, outline caching, font loading).
  • Before committing optimizations — run cold first so cache impact is visible.
  • Comparing before/after for a specific renderer (PNG vs SVG vs PDF).

Prerequisites

  • Rust toolchain with --release.
  • KaTeX TTFs in fonts/ (committed in-repo; 20 .ttf files).
  • (Optional) CJK/emoji system fonts for the full CJK/emoji categories.

Commands

1. Full 100-formula benchmark

cargo test -p ratex-render --test bench_render --release -- --ignored --nocapture

Renders 100 formulas across 6 categories (math, complex, matrix, cjk, emoji, chem) in PNG, SVG, SVG-standalone, and PDF. Prints per-category averages and overall throughput (formulas/sec).

Warmup: 1 iteration per formula. Measurement: 3 iterations, averaged.

2. Font cache cold vs hot timing

cargo test -p ratex-render --test font_cache_timing -- --ignored --nocapture

Measures 3 formulas: first pass (cold — disk I/O for font loading), second pass (hot — OnceLock cache hits). Reports speedup ratio.

3. Render phase breakdown

cargo test -p ratex-render --test phase_breakdown --release -- --ignored --nocapture

Per-formula breakdown: parse+layout vs render phase timing. 6 formulas including CJK and emoji cases. Warmup: 3 iterations, Measurement: 10 iterations, averaged.

4. Compare before/after (quick diff)

Useful when optimizing a specific path:

# Before (on main)
cargo test -p ratex-render --test phase_breakdown --release -- --ignored --nocapture 2>&1 | tee /tmp/before.txt

# After (on branch)
cargo test -p ratex-render --test phase_breakdown --release -- --ignored --nocapture 2>&1 | tee /tmp/after.txt

# Diff the tables
diff /tmp/before.txt /tmp/after.txt

Quick checklist

  1. Run cargo test -p ratex-render --test bench_render --release -- --ignored --nocapture for the full picture.
  2. For quick iteration, use phase_breakdown (faster, 6 formulas).
  3. For font-loading specific changes, use font_cache_timing.
  4. Always use --release — debug builds are 10-50× slower and meaningless for perf comparison.