It's true in programming and its true everywhere else: don't optimize before you find the real bottlenecks. Otherwise you'll have a 1% gain but all sorts of unforeseen consequences start popping up.