You don't really know beforehand which way will be faster. But that kind of "optimisation" is not very useful and can be counterproductive. If you have a program that doesn't run fast enough you need to profile it to see where it is spending most of its time. That's where you can make the most difference.
So in general you should use whichever method makes the algorithm more clear.
Although, now that I think about it, if you just want to copy the data from one array to the other then memcpy is perhaps fastest.