llama.cpp

llama : make quantize example up to 2.7x faster

#3115

Merged