APEX Compute - a zero-dependency C++20 engine built from scratch to execute cache-tiled, explicitly vectorized matrix multiplication (GEMM) on ARM and x86 CPUs.

It bridges the gap between raw compute performance and verified correctness by coupling bare-metal SIMD kernels with automated benchmarking (apex-bench) and differential validation suites (apex-validate).

An optional natural-language intelligence layer using Gemma 4 that translates user intent into structured hardware run plans.

Check out the repository:- apex_compute