llama.cpp

f266259a - Speedup the AVX-512 implementation of ggml_vec_dot_q4_0() (#933)