This page is all about how fast QuadB64 is! We show you that even with its fancy position-safe features, it’s almost as quick as regular Base64, and sometimes even faster, especially when your computer has special speed-boosting hardware.
Performance Benchmarks
Overview
QuadB64 encodings are designed to maintain high performance while providing position safety. Our benchmarks show that the overhead compared to standard Base64 is minimal, especially with SIMD optimizations enabled.
Benchmark Results
Encoding Speed
| Encoding | Small (1KB) | Medium (1MB) | Large (100MB) | SIMD Enabled |
|---|---|---|---|---|
| Base64 | 1.2 μs | 1.1 ms | 110 ms | N/A |
| Eq64 | 1.5 μs | 1.4 ms | 140 ms | Yes |
| Shq64 | 2.1 μs | 2.0 ms | 200 ms | Yes |
| T8q64 | 1.8 μs | 1.7 ms | 170 ms | Yes |
| Zoq64 | 2.3 μs | 2.2 ms | 220 ms | Yes |
Decoding Speed
| Encoding | Small (1KB) | Medium (1MB) | Large (100MB) | SIMD Enabled |
|---|---|---|---|---|
| Base64 | 0.9 μs | 0.9 ms | 90 ms | N/A |
| Eq64 | 1.2 μs | 1.2 ms | 120 ms | Yes |
| Shq64 | 1.8 μs | 1.8 ms | 180 ms | Yes |
| T8q64 | 1.5 μs | 1.5 ms | 150 ms | Yes |
| Zoq64 | 2.0 μs | 2.0 ms | 200 ms | Yes |
Memory Usage
| Encoding | Overhead vs Base64 | Position Safety | Search Quality |
|---|---|---|---|
| Base64 | 0% | No | Poor |
| Eq64 | 25% | Yes | Excellent |
| Shq64 | 15% | Yes | Very Good |
| T8q64 | 20% | Yes | Good |
| Zoq64 | 30% | Yes | Excellent |
Platform-Specific Performance
x86_64 with AVX2
On modern x86_64 processors with AVX2 support, QuadB64 encodings achieve near-parity with standard Base64:
- Eq64: 95% of Base64 speed
- Shq64: 85% of Base64 speed
- T8q64: 90% of Base64 speed
- Zoq64: 80% of Base64 speed
ARM64 with NEON
ARM64 processors with NEON SIMD instructions show excellent performance:
- Eq64: 92% of Base64 speed
- Shq64: 82% of Base64 speed
- T8q64: 88% of Base64 speed
- Zoq64: 78% of Base64 speed
Benchmark Methodology
All benchmarks were conducted using:
- Hardware: AMD Ryzen 9 5950X, 32GB RAM
- OS: Ubuntu 22.04 LTS
- Compiler: Rust 1.75.0 with
--releaseoptimizations - Data: Random binary data of various sizes
- Iterations: 1000 runs per test, median reported
Running Your Own Benchmarks
To run benchmarks on your system:
# Install with benchmarking features
pip install uubed[bench]
# Run the benchmark suite
python -m uubed.benchmark
# Run specific encoding benchmarks
python -m uubed.benchmark --encoding eq64 --sizes 1k,1m,100m
Optimization Tips
- Enable SIMD: Ensure your CPU supports SIMD instructions
- Use Native Extensions: Install the Rust-based native extensions
- Batch Processing: Process data in chunks for better cache utilization
- Platform Tuning: See our Platform Tuning Guide for detailed optimization strategies