How would you describe them?
Arm NEON kernel for llama.cpp tiled K-quant matmul - benchmark study, patch and tools
