llama.cpp

llama : fix integer overflow during quantization

#4284

Merged