zig-crypto-bench

Crypto benchmarks: Zig vs Rust performance comparison.

Overview

Cryptographic performance benchmarks across Zig stdlib and Rust implementations with hardware acceleration.

Supported platforms: Linux and macOS (x86_64, aarch64)

Benchmarked Algorithms

• SHA-256: Secure Hash Algorithm 256-bit

  • Zig: std.crypto.hash.sha2.Sha256
  • Rust: sha2 crate with hardware acceleration

• SHA-512: Secure Hash Algorithm 512-bit

  • Zig: std.crypto.hash.sha2.Sha512 *
  • Rust: sha2 crate with hardware acceleration

  * Zig's SHA-512 currently lacks hardware acceleration (#22266), resulting in expected performance differences.

Results

Run benchmarks to generate timestamped results in results/ directory:

zig build bench -- --save-results

Installation

Prerequisites

• Zig 0.14.1 or later
• Rust toolchain (for building the comparison library)

Build Instructions

git clone https://github.com/yourusername/zig-crypto-bench
cd zig-crypto-bench
zig build bench

Usage

# Run all benchmarks with default settings (500 iterations, 50 warmup)
zig build bench

# Save timestamped results to results/ directory for contribution
zig build bench -- --save-results

# Run with custom iteration counts
zig build bench -- --iterations 2000 --warmup 200

# Generate JSON output for analysis
zig build bench -- --json > results.json

# Filter specific tests by size
zig build bench -- --filter 1MB

# Combined: save results with custom settings
zig build bench -- --save-results --iterations 1000 --warmup 100

Features

• Adaptive iteration counts with auto-scaling for large inputs
• Warmup phase to stabilize CPU performance
• Auto-batching for accurate timing of fast operations
• Correctness verification via bitwise output comparison
• Transparent build configuration and optimization flags
• Statistical analysis
• CPU frequency governor detection and throttling warnings
• Framework versioning with git commit tracking in results
• Two measurement modes: FFI (both use C ABI for fair comparison, default) and native (Zig direct/Rust FFI)

Contributing

Contributing Results

Hardware Results: Help expand hardware coverage by contributing benchmark results from your system. Run with --save-results and submit the generated markdown files from the results/ directory.

Open Issues: Check the issues for tasks ranging from small improvements to major features.

Development Conventions:

• Test functions must print their title at the start using std.debug.print("\n✓ Running: Test Name\n", .{}); (This is idiomatic in Zig as the test runner lacks built-in verbose output)