Where to learn about threading logic in llama.cpp

[Nick-infinity]

Hello, I am trying to understand how multithreading works in llamma.cpp/gmml and where is the control point. While looking at the operator code, I can see only the microkernel. I assume this microkernel is called with N threads to perform chunking and parallelization. Can someone please help me in better understanding how multiple threads are controlled within the operator and who is the owner/ caller of these N thread calls?

Thanks,