llama.cpp

llama : do a warm-up eval at start for better timings

#1824

Merged