If the need for floating-point is only internal to your system program, you have a lot of options on how to compute mathematic functions etc. There are many tricks that can be exploited if you know how or employ someone that knows how.

However, if you take dwnloaded floating point, your hands are much more tied to standard approaches. Standards usually sacrifice computational effiency but they make a safety net - allowing you to get results without understanding all the math.

-Jay C. Box