llama.cpp

e25fb4b1 - ggml : use `uint8x16_t` return type for `ggml_vqtbl1q_u8` (#5894)