> No one here is saying that extended precision is bad. It is very > important for correct results for some algorithms. That is why the IEEE > FP standards require its presense. That is wrong, this is not a required element of the standard