c++ comparing two floating point values

Question

I'm wondering what is the difference for comparing two double between this two manner :

double a1 = ...;
double a2 = ....;

1. fabs(a1-a2) < epsilon
2. (fabs(a1-a2)/a2) < epsilon

Is there a prefering way to do that ?

thanks

Answer 1

This article answers your question quite thoroughly, I think. You might want to skip ahead to the section "Epsilon comparisons".