What every computer scientist should know about floating-point arithmetic

[TriinT]

http://docs.sun.com/source/806-3568/ncg_goldberg.html

[mhansen]

94 Pages? I'll sum it up for you here:

Rounding error happens. It's worse if you do add/subtract a very big number and a very small number.

Rounding error increases as you do more operations.

Depending on your platform, you may need to accomodate NaN's and positive/negative infinity.

[jacquesm]

If you really need floating point you don't understand your problem.

Old programmers joke from the time when cpus didn't have floating point co-processors.

It's possible, but it certainly is tedious.