4 ms·

80-bit floats are just annoying and non-portable, and doesn't even help in the case mentioned. But double-double representation is nice and portable. Also, if

• dkarapetyan 12y ago
  Why is needing more significant digits than the hardware can provide doing something wrong? At the end of day the tools should be subordinate to the problem at hand and not the other way around. You're arguing for the primacy of tools over the actual problem being solved. That sounds backwards.