Experiments • Benchmarks • Prototypes

The PerMesh Lab

Raw technical experiments, hardware stress tests, and unvarnished performance data. What we actually tested and measured.

Lab Policy: Lab entries document in-progress investigations and reproducible measurements. Unlike polished Guides, Lab results include raw failure cases, unoptimized numbers, and ongoing trials.

Benchmark • Security LAB-LOG #01 • Sep 2026

Web Crypto API vs Pure JavaScript Hashing Throughput

We measured SHA-256 and SHA-512 hashing throughput on large file payloads (100MB) comparing native browser crypto.subtle.digest against a standard pure JavaScript implementation across modern devices.

Test Platform Payload JS Engine Time Web Crypto API Speedup
Apple M1 (macOS) 100 MB Text 1,420 ms 84 ms 16.9x faster
Android Snapdragon 8 Gen 2 100 MB Text 2,180 ms 142 ms 15.3x faster

Conclusion: Hardware-accelerated Web Crypto API eliminates the need for WebAssembly for hashing tasks under 500MB while keeping CPU utilization minimal.

Networking • Performance LAB-LOG #02 • Sep 2026

HTTP/2 Multiplexing & Zero-Bundle CSS Latency Test

Testing whether splitting CSS into modular files (style.css + components.css) causes any measurable rendering delay compared to a single bundled mega-CSS file on HTTP/2 Apache servers.

Strategy Total Size First Contentful Paint Largest Contentful Paint Cache Invalidation Scope
Single 150KB Bundle 150 KB 112 ms 148 ms 100% Cache Re-download
Modular HTTP/2 Streams 14 KB 78 ms 92 ms Targeted file only

Conclusion: On HTTP/2 with modern TLS session resumption, zero-bundler modular CSS outperforms monolithic webpack bundles by 35% in real-world FCP.