llama.cpp

ggml-hexagon: optimize activation function

#18393

Merged