llama.cpp

a9559bf7 - ggml : workaround for missing _mm256_setr_m128i in GCC < 8 in k_quants.c (#2405)