Why would you go to all that trouble when it would be easier just to remember 3.14? Those three digits alone will give you more accuracy than any real life measurements are likely to have in the first place.
I use that when I'm doing my moon shuttle gig. It helps cut down on the course corrections.
It's one of those things you need as a programmer if you find yourself working on a small system (like embedded stuff) where the cost of implemented floating point math is just too great, either in extra circuitry or in extra memory. Under conditions where the programmer is confined to using fixed-point (scaled integer) math, calculations involving transcendental numbers can still be done using integer multiply/divide sequences. Need the area for a circle? Have radius? It's a simple matter to perform the integer multiplication: V = R x R x 355, and then the integer divide A = V / 113.
People using computers running modern CPUs and floating-point co-processors don't worry about stuff like this. People writing for 8-bit and 16-bit embedded processors have to pay attention to these things.
Yes, there are still systems that use low-end (inexpensive) processors, and you'd be amazed at the performance you can get from an 8051 or 6809 running at 12MHz if it's not running Windows or some other heavyweight OS.
