llama.cpp

184c694f - ggml-cpu: optimize ggml_vec_dot_bf16 for s390x (#19399)