SoA vs AoS Performance Benchmarks

Benchmarks comparing Structure of Arrays (SoA) vs Array of Structures (AoS) performance in Rust.

Running Benchmarks

Run all benchmarks

cargo bench

Run with native CPU optimizations (recommended)

RUSTFLAGS='-C target-cpu=native' cargo bench

Run specific benchmark files

cargo bench --bench my_benchmark
cargo bench --bench auto_optimisation_benchmark

Run selective benchmarks by name

# Run only AoS benchmarks
cargo bench "AoS"

# Run only SoA benchmarks
cargo bench "SoA"

# Run only SIMD benchmarks
cargo bench "SIMD"

# Run SIMD width comparison
cargo bench "SIMD x"

# Run specific benchmark
cargo bench "SoA Update (zip)"

Benchmark Files

• my_benchmark.rs: Core SoA vs AoS comparison including SIMD variants
• auto_optimisation_benchmark.rs: Tests compiler optimization behavior

What's Being Tested

• AoS Update: Traditional struct-based particle updates
• SoA Update: Array-based particle updates (better cache locality)
• SoA Update (zip): Idiomatic Rust with iterator zipping
• SIMD variants: Manual SIMD implementations at different widths (x4, x8, x16, x32)