中文版

Automated performance comparison of Lua and JavaScript engines. Build from source, run standardized benchmarks, and generate visual reports — all with a single command.

Planned: Luau, JavaScriptCore (JSC)

git clone https://github.com/xjdrew/lua-js-benchmark.git
cd lua-js-benchmark

# One command does everything: check deps → download source → build → benchmark → report
make all

Or step by step:

make setup    # Download and build all engines
make bench    # Run benchmarks (default: 5 runs, 2 warmup)
make report   # Generate Markdown + HTML reports

Reports are saved to results/<timestamp>/:

• report.md — Markdown tables, viewable on GitHub
• report.html — Interactive charts (Chart.js), open in any browser

Each benchmark has both a Lua and a JavaScript implementation with verified identical output. See benchmarks/README.md for detailed descriptions, sources, and conventions.

Platforms: Linux (x86_64, aarch64), macOS (x86_64, Apple Silicon)

Required tools:

Optional (for V8 only): g++/clang++, ninja

Run make setup to automatically check dependencies. Missing tools are reported with platform-specific install commands.

sudo apt update && sudo apt install -y build-essential git curl python3 python3-venv time ninja-build

sudo dnf install -y gcc gcc-c++ make git curl python3 python3-pip time ninja-build

sudo pacman -S --needed base-devel git curl python python-pip time ninja

xcode-select --install
brew install bash git curl python3 ninja

# Skip V8 (avoids large download)
make setup ENGINES="lua luajit quickjs"
make bench ENGINES="lua luajit quickjs"

make bench CATEGORY=compute
make bench CATEGORY=string

make bench RUNS=10 WARMUP=3

make download-lua       # Retry Lua source download
make download-v8        # Retry V8 source download
make build-lua          # Retry Lua compilation
make build-v8           # Retry V8 compilation

You can also manually place source code in .build/src/ and run the corresponding build command.

make clean              # Remove .build/ and results/
make rebuild            # Clean + setup

All comparisons use Lua 5.4 as the baseline (1.00x). Values below 1.00x mean faster than Lua; above means slower.

The HTML report includes:

• Stacked bar chart — Overall performance. Each bar is an engine, segments are categories. Shortest bar = best engine.
• Category comparison — Grouped bars per category with 1.0x baseline.
• Radar chart — Multi-dimensional view of engine strengths.
• Memory usage — Peak RSS comparison.
• Detailed table — Every benchmark with absolute times and relative multipliers.

1. Create scripts/build_<engine>.sh (download + compile)
2. Add download logic to scripts/download_sources.sh
3. Register in runner/config.sh: binary path, language, and runtime args
4. Run make setup && make bench

1. Add the Lua script to benchmarks/lua/<category>/<name>.lua
2. Add the JS script to benchmarks/js/<js-category>/<name>.js
3. Ensure both produce identical output for the same input
4. Generate expected output: lua benchmarks/lua/<category>/<name>.lua > benchmarks/expected/<category>/<name>.txt
5. Run make bench — new benchmarks are discovered automatically

lua-js-benchmark/
├── Makefile                # Top-level entry point
├── scripts/                # Environment setup and engine build scripts
│   ├── setup.sh            # Main setup orchestrator
│   ├── check_deps.sh       # Dependency checker
│   ├── detect_platform.sh  # OS/arch detection
│   ├── download_sources.sh # Source code downloader
│   ├── build_lua.sh        # Lua build script
│   ├── build_luajit.sh     # LuaJIT build script
│   ├── build_quickjs.sh    # QuickJS build script
│   └── build_v8.sh         # V8 build script (stripped: no wasm/ICU/debug)
├── benchmarks/
│   ├── lua/                # Lua benchmark scripts
│   ├── js/                 # JavaScript benchmark scripts
│   └── expected/           # Expected output for correctness checks
├── runner/
│   ├── config.sh           # Engine registry and parameters
│   ├── measure.sh          # Single-run measurement wrapper
│   └── run.sh              # Benchmark runner
├── report/
│   ├── generate.py         # Report generator
│   └── templates/          # HTML template with Chart.js
├── docs/                   # Project planning documents
├── .build/                 # Downloaded sources and binaries (git ignored)
└── results/                # Benchmark results and reports (git ignored)

• Project Plan — Design decisions, architecture, and rationale
• Implementation Plan — Phase-by-phase progress tracker

See CONTRIBUTING.md for guidelines on how to contribute.

This project is licensed under the MIT License. See LICENSE for details.