Here's a challenge for floating point lovers.

[Doug Gwyn]

In article <2855 at charon.cwi.nl> dik at cwi.nl (Dik T. Winter) writes:
>In article <101 at tdatirv.UUCP> sarima at tdatirv.UUCP (Stanley Friesen) writes:
> > #define FLT_ROUNDS (__flt_rounds)
>And right they are because it can change at runtime.

However, the C standard does not contemplate rounding mode changing
during execution of a strictly conforming program.  That stuff is
an IEEE floating-point extension that need not be supported by a
conforming implementation.  (I'm trying to resist my usual slam at
IEEE floating point.)