llama.cpp b9459 Update Introduces Metal GLU Kernel Templates for f16/f32

Core summary

One sentence to understand this update

llama.cpp release b9459 updates Metal GLU kernels to use templates for f16/f32 support, optimizing memory bandwidth by using native tensor types.

Impact & opportunity

What this could mean

This update allows llama.cpp to run more efficiently on Metal-enabled hardware, potentially improving performance and memory usage for local LLMs.