CHRIS MCBRIEN had something deep to write about C/Set++


CM> My system tells me the value is 2.639996E+01.  Now that's not a
CM> huge difference, I admit, but it's enough to really screw up my
CM> program.

Considering that a float is only 32 bits long, and only 23 or 24 (I
can't remember which, and am too lazy to look it up) are used to
represent the mantissa, you only have about 7 digits of accuracy.  If
you need more, use a double instead of a float.  This isn't a flaw of
OS/2, but because 26.4 cannot be exactly represented by a float.  It
can't be exactly represented by a double, either, but the error is
less. See example below:

#include 
main(){
float temp=26.4;
double dtm=26.4;
printf("\nfloat: %25.10f  %25.10e\n",temp,temp);
printf("\ndouble:%25.10lf  %25.10le\n",dtm,dtm);
}

result: (Note, this was a dos program, not OS/2, but both use the same
math)
 float:             26.3999996185          2.6399999619e+001

 double:            26.4000000000          2.6400000000e+001

 * KWQ/2 1.2g * So easy, a child can do it. Child sold seperately


--- QScan v1.12b / 01-0093