On Tuesday, 26 November 2013 at 21:11:25 UTC, David Nadlinger
wrote:
IEEE 754 includes infinity as an actual value, contrary to the
usual definition of real numbers in mathematics.
A bit late, but AFAIK inf represents either overflow or N/0...
And 0 represents either underflow or zero. Computations on those
values should be conservative unless you trap overflow/underflow
exceptions and handle those as special cases. You want to
preserve overflow... Then you have the denormal numbers
(underflow where you retain some digits). It is tempting to think
of FP as real numbers, but they are not, of course, so libraries
should IMO be conservative.
