llama.cpp

fix(avx): workaround for missing _mm256_set_m128i in GCC < 8

#1638

Merged