Hi, thanks, I initially understood that, but I gave a non parallelizable example to show that there sometimes it is serialized by design, but I agree that in theory "real world" examples should be more amenable to parallelization.

Looking forward to seeing how the hardware handles the cost of communication between cores.