Maybe you could try with ArbitraryPrecisionFloat package (browse
http://www.squeaksource.com).
But you should then know the Number of bits to use in advance...
Or try a package that can deal with AlgebraicNumber, maybe MathMorph
would do that?
In Smalltalk, ideas are programmed so fast that it's fun...
Gregory Hinton a écrit :
Hmmmm, good catch. I had overlooked the finite precision of floating
point.
Looks like it's accurate up to "75 nthFibonacci" but thereafter diverges
from reality.
This is why I prefer mathematics to computer science: mathematicians can
assume the existence of irrational numbers. ;)
| 