[CLN-list] floating point underflow - a possible explanation

[Richard B. Kreckel]

Michael Schuster wrote:
> So once I have cl_SF in my system, it can be anywhere. So now I changed 
> default constructor to 
> 	Real():cl_F(cl_float(cl_I(0),float_format(_PRECISION_CLN_NEWMAT))) {};
> and everthing seems ok now. 

What you have observed is intentional: There is no point keeping extra 
precision once it's been poisoned with an operand that has fewer digits.

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