llama.cpp

llama-bench: add --n-cpu-moe support

#15952

Merged