As a side note:

Fibonacci is one of these theoretical problems which need large computation where implementation matters.

But in real life you can approximate the result with a certain accuracy.

So why should I need more than 10 digits of fib (100) if I ever need fib(100) ?