29 Dec
2007
29 Dec
'07
9:10 p.m.
Alexei Sheplyakov wrote:
// approximation for smallest number so that 1+tol > 1 numeric tol=ex_to<numeric>(3*pow(ex(10),ex(-(Digits-2))));
Not really.
Let me just add that plain CLN does come with a better function for that. From the manual: `cl_F float_epsilon (float_format_t f)' Returns the smallest floating point number e > 0 such that 1+e != 1. -richy. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>