OSDev.org

Hi all,

had a bit of time on my hands and so I wrote something i wanted to implement for a while now, namely a program to calculate the roots of cubic expressions (by Cardano's method). Now I've written it, I am not entirely sure if I'm doing everything right. I added a check for the results at the end, and I regularly get results on the order of 1e-16. Is that entirely down to my non-optimized programming? Or did I get something wrong?

Double-precision values can accurately store around 15 decimal digits, so that sounds like normal rounding error.

By default, compilers don't change the order of floating-point operations while optimizing, so you may be able to improve the rounding error by changing how you calculate intermediate values.