llama.cpp

a9e8a9a0 - ggml : fix arch check in bf16_to_fp32 (#10164)