Also, in my experience, maximum performance for most algorithms is not a clear cut line that I fall short of for lack of motivation. I see that maximum as a rather fuzzy unknown, approaching it is has diminishing returns on the time spent optimizing.