llama.cpp

6eeaeba1 - cmake: use 1 more thread for non-ggml in CI (#8740)