Performance Benchmarks

This document provides comprehensive performance benchmarks for gomjml compared to other MJML implementations.

Overview

The benchmark suite compares three MJML compilation tools using a realistic email template (Austin layout from mjml.io):

The benchmarking script (bench-austin.sh) uses a two-pass approach to ensure accurate measurements:

Pass 1: Clean timing measurements with no monitoring overhead (50 iterations each)
Pass 2: Resource monitoring for CPU and memory usage (50 iterations each)

This separation prevents the monitoring overhead from affecting timing accuracy while still collecting comprehensive resource usage statistics.

Tool	50x Total (ms)	Avg (ms)	Max RAM (MB)	Avg CPU (%)
gomjml	162	3	2	0
mrml	96	1	1	0
mjml (JS)	13530	270	90	20.9

Native Performance: Both Go (gomjml) and Rust (mrml) implementations significantly outperform the JavaScript version
Memory Efficiency: Native implementations use minimal memory (1-2MB) vs JavaScript (90MB)
CPU Usage: Native implementations have negligible CPU overhead vs JavaScript (20.9% average)
Production Readiness: For high-throughput email generation, native implementations are essential

To run the benchmarks yourself:

# Run with ASCII table output
./bench-austin.sh

# Run with markdown table output
./bench-austin.sh --markdown

# Get help
./bench-austin.sh --help

The benchmarking script has evolved to address timing accuracy issues:

Initial Implementation: Simple timing with resource monitoring caused measurement interference
Problem: Monitoring overhead affected timing results (2ms→22ms degradation)
Solution: Two-pass methodology separating timing from resource monitoring
Enhancement: Added markdown output support for documentation integration

This ensures both accurate timing measurements and comprehensive resource usage statistics.