Ridiculous fast optimized implementation of matrix multiply on NEON ARM CPUs - View it on GitHub