On Thu, 10 Mar 2016, Volker Braun wrote: (Not Sage-specific)
IEEE doesn't raise errors because most common CPUs have no special support for that.
I have also read another explanation: A user may want to continue computation (maybe in parallel) and only as a last stage drop out those parts of the result that ended with NaN.
Years ago Borland Delphi had an option to throw exception for zero division. And now GCC seems to have option -mfp-exceptions.
-- Jori Mäntysalo
