A modular benchmark for C++ optimisation — memory layout, SIMD, lock-free concurrency, and latency-spike avoidance. An evolving living reference.
Overview
A sandbox for the techniques that make audio code fast and predictable. Each module isolates a concern — data layout, vectorisation, threading — so the impact of a given optimisation can be measured rather than guessed at.
Focus areas
Cache-friendly memory layout, SIMD vectorisation, lock-free concurrency between the audio thread and the rest of the system, and the discipline of avoiding latency spikes in the real-time path. It's maintained as a living reference that grows as new techniques prove out.