llama.cpp

ggml : spread compute across threads in chunks

#1507

Open