llama.cpp

47226716 - vulkan: improve mul_mat_vec_iq1_s speed (#17874)