llama.cpp

6958d413 - sampling : check backend support during init