BUG:
data vectors were reloading the entire file each time they got new data,
rather than shifting them. And shifting didn't use memmove.

These changes should speed up real time significantly.