llama.cpp

cb79c2e7 - ggml: don't include arm_neon.h when using CUDA 12 with ARM Neon (ggml/1187)