Re: IEEE floating point support for guile?

[Mikael Djurfeldt]

Mikael Djurfeldt <address@hidden> writes:

> At any rate, it seems to me that also the case when floating-point
> exceptions aren't generated can be useful.

I've now spoken to a numerics guy.  He agrees that generating
exceptions at division by zero would be a serious limitation.

He confirms my own experience that it is common that you do
computations where parts of the result are NaN:s or infinite values.

He agreed that having the option to generate numerical exceptions
could be useful during initial debugging of the program, but thinks
that no exceptions should be the default behaviour.