Given that there is a split command (at least, on most Unixes), I would be very strongly inclined to try to use it, hoping that its implementation is most efficient.   (In any case, it is an existing implementation of a classic Thing That Is Already Done.™)

“Splitting a file” is never a thing that should call for recursion:   all that you’re really doing is reading from one file and switching from one output-file to the next one at specified intervals.   I’ve seen lots of ways to do that, probably the most-elaborate ones (not in Perl ...) using memory-mapped files to actually exploit the virtual-memory subsystem’s I/O capabilities as the means of reading from the target and getting the data (in one step) where it needs to go.

Still, in my mind, it all comes back to the same thing:   this is A Classic Thing That Has Already Been Done.™   Search for an existing tool that you can rely-upon, and use it, to avoid having to write-and-debug “yet another” piece of software to do such a trivial task.   Surely you can find one that will meet your project’s performance expectations.