although compiled code looks MUCH less optimal than this, I haven't noticed any performance increase :-/