[mgaunard]

Achieving exact or correctly-rounded results is much more work than that.

You can look at the crlibm papers for example.