llama.cpp

9d1e0361 - cuda : fix -Wtautological-compare in fattn vec dispatch (#30519)