Maybe some kind of opportunistic prefetching if you have regular deadlines and can expect a regular complex problem deadline. But maybe concurrent threads would be easier, so this could make sense only in a super limited embedded environment.