It looks like upstream work is underway: https://github.com/ggml-org/llama.cpp/issues/29058

The sheer number of geniuses working so fast on this repo never ceases to amaze me.