simd-optimization

(WIP) Ultra-fast, SIMD-accelerated Predictive Coding (PC) model library in bare-metal C++

A hardware-accelerated, zero-heap C++ Graph Neural Network (GNN) inference engine for edge deployments. Utilizes 256-bit SIMD (AVX2/FMA) to achieve ultra-low latency (~4.1µs) execution. Features a local microservice daemon with JSON IPC and an in-line physics validation layer.