Rounding Error

[Eric Mercer]

I have run into an interesting situation using the following
configuration:

 * gcc version egcs-2.90.27 980315 (egcs-1.0.2 release) 
   shipped with RedHat 5.1
 * Pentium II 266 Mhz running RedHat 5.1

When I compile and run the following snippet of code:

#include <iostream.h>
#include <math.h>

main( int argc, char *argv[] ) {

  double tolerance = 0.001;
  cout << endl << "ceil(1/" << tolerance << ") -> " << ceil( 1.0/tolerance);
  cout << endl << "floor(1/" << tolerance << ") -> "<< floor(1.0/tolerance );
  cout << endl << "(int)(1/" << tolerance << ") -> " 
       << (int)((double)1.0/tolerance );
  cout << endl;
  
  tolerance = 0.01;
  cout << endl << "ceil(1/" << tolerance << ") -> " << ceil( 1.0/tolerance);
  cout << endl << "floor(1/" << tolerance << ") -> "<< floor(1.0/tolerance );
  cout << endl << "(int)(1/" << tolerance << ") -> " 
       << (int)((double)1.0/tolerance );
  cout << endl;

  tolerance = 0.1;
  cout << endl << "ceil(1/" << tolerance << ") -> " << ceil( 1.0/tolerance);
  cout << endl << "floor(1/" << tolerance << ") -> "<< floor(1.0/tolerance );
  cout << endl << "(int)(1/" << tolerance << ") -> " 
       << (int)((double)1.0/tolerance );
  cout << endl;

}

I get the following incorrect results:

ceil(1/0.001) -> 1000
floor(1/0.001) -> 1000
(int)(1/0.001) -> 999

ceil(1/0.01) -> 100
floor(1/0.01) -> 100
(int)(1/0.01) -> 99

ceil(1/0.1) -> 10
floor(1/0.1) -> 10
(int)(1/0.1) -> 9

Now, Is this a problem in glibc, egcs, or the Pentium II hardware?

egm