It's not a problem with rounding so much as a problem storing the number in the first place. Double precision floats have roughly 16 decimal digits of precision.

12314212412124124124.10623  (Your number)
|------16------|
12314212412124123136.00     (The output)
[download]

You'll need something that can support more precision, such as Math::BigFloat or build of Perl that supports quad precision floats.