[CLN-list] Floating point overflow for cl_R ln (const cl_R& x)

[Richard B. Kreckel]

Hi!

On Sat, 8 Jan 2005, Johnny Fundalewicz wrote:
> I'm getting some floating point overflow errors when using logarithmic
> functionc l_R ln (const cl_R& x) with big Numbers ( > 20 Stellen).
>
>
> My code looks like this:
>
> --------------------------------
> #include <cln/cln.h>
>
> int main(){
>
> cln::cl_R number_one =
> "41234123412341234123157311812341234124312359123412341";
>
> cln::cl_R number_two = cln::ln (number_one);
>
> std::cout << number_two << std::endl;
>
> return 0;
> }
> --------------------------------
>
>
> I compiled with gcc-Version 3.3.5 (Debian 1:3.3.5-4):
>
> 	g++ -g -o test -lcln test.cpp
>
>
> Have I done a mistake or whats the reason why I can't compute those
> values - Is the number too large?

What happens is that CLN figures that the result is going to be a float
and converts the result to a cl_F of precision default_float_format before
calling the function ln(cl_F).  But the exponent is too large and the
conversion step triggers the overflow.

Just try increasing the precision and your code work.  You can easily
check this by printing most_positive_float(default_float_format).

Regards
  -richy.
-- 
Richard B. Kreckel
<http://www.ginac.de/~kreckel/>