What is the most effective way to duplicate a BITMAP?
blit() does the job of course, but would memcpy or such be more efficient?

For the best speed, use the SSE registers... (assuming Intel arch)

Does blit call memcpy internally? Probably checks what type of bitmap and calls the optimal routine.

I think so.

Either way - it has to be blit, otherwise pointers in the bitmap struct itself are not set properly.