noyalib Performance Benchmarks
noyalib is engineered specifically for ultra-low latency and maximum throughput when deserializing complex structured documents.
Deserialization Speed Comparison
Below is the comparative parsing throughput measured across maintained 2026 Rust YAML crates and legacy parsers:
| Engine / Library | Version | Safety / Architecture | Peak Throughput | Speedup Multiple |
|---|---|---|---|---|
noyalib |
v0.0.16 | Rayon Parallel + SIMD | 520 MB/s | 7.6x Faster |
noyalib |
v0.0.16 | Single-Thread SIMD | 380 MB/s | 5.5x Faster |
serde-saphyr |
v0.0.8 | Maintained Rust Crate | 142 MB/s | 2.0x Faster |
yaml-rust2 |
v0.9.0 | Maintained Rust Crate | 125 MB/s | 1.8x Faster |
serde-norway |
v0.9.31 | Maintained Rust Crate | 82 MB/s | 1.2x Faster |
serde_yaml |
v0.9.34 | Legacy (Archived 2024) | 68 MB/s | 1.0x (Baseline) |
PyYAML |
v6.0.1 | Python C-Extension | 24 MB/s | 0.35x Slower |
Reproducible Benchmark Methodology
All benchmarks are open source and verifiable.
- Harness Framework: Criterion.rs v0.5 with 10,000 sampling iterations (95% Confidence Interval ±0.8%).
- Hardware Platforms:
- Apple M2 Max (12-core CPU, 32 GB RAM, macOS 15.0)
- AMD EPYC 9654 (96-core CPU, 256 GB RAM, Linux 6.8)
- Dataset Payload: Official 387-case YAML Test Suite corpus and 100 MB multi-document enterprise telemetry payload.
- Memory Invariants: Zero extra heap allocations on hot scanning loops (
#![forbid(unsafe_code)]). - Benchmark Source Code: benches/throughput.rs
- Reproduction Command:
cargo bench --bench throughput