FastAudioProcess Reference Specification

Detailed JNI contracts, fallbacks, and execution rules for FastAudioProcess.

1. CPU & SIMD Feature Model

⚡ JDK Vector API (SIMD) — Used for high-performance JVM-level vectorization (e.g., RMS calculation, normalize, mixChannels).
🚀 AVX2 / SSE4.2 — detected via CPUID for native routines. Enables 32-byte and 16-byte vector ops.
🔄 Fallback rule: JDK Vector API / Native AVX2 → SSE4.2 → scalar.

🚫 Zero-Copy — All operations use direct memory pinning (GetPrimitiveArrayCritical) or direct buffers for zero copy between Java and native layers.
📏 Unaligned Access — Safe on all byte boundaries.
🔒 Thread-Safety — All static native and Java processing methods are thread-safe.

⚡ Direct Memory Pinning — No implicit copies are made by the JNI bridge.
📦 No Allocation — All operations work on pre-allocated Java arrays or buffers.
⏰ Critical Sections — Native calls minimize blocking to prevent GC impact.

detectPitchNative(float[] samples, int sampleRate): Uses JNI autocorrelation mapping to estimate the fundamental frequency. Returns float pitch in Hz, or 0.0f if unvoiced.
pitchShiftNative(float[] samples, float semitones, int sampleRate): Modifies pitch in-place using a zero-copy time-domain Overlap-Add algorithm in native C++, preserving duration.

logMelSpectrogram(float[] samples, ...): Computes linear FFT bins and maps them onto log Mel-scale spectrogram bands.
FrameChunker: Overlapping sliding-window chunker for stream block segmentation.
generateWaveformPoints(float[] samples, int targetPoints): Mathematical downsampling of audio signals to a fixed number of peak values for visualization/drawing.
getFramePeak(float[] samples): Calculates the maximum absolute peak value of a single audio frame (useful for real-time stream level meters).

Platform	Status
Windows 10/11 (x64)	✅ Fully Supported

Part of the FastJava Ecosystem — Making the JVM faster.

Made with ⚡ by Andre Stubbe