From: Ian Dall <Ian.Dall@dsto.defence.gov.au>

> To read a double value using scanf you need a format of "%lg", otherwise
> it writes a single into the address of a double...

Of course, using a correct program eliminates many errors...  I
just retried it with %lg and it looks better, but gcc and scanf
still do not appear to agree on the equality of that constant.

> As for doubles less than DBL_MIN, there is certainly support for denormalized
> numbers in the floating point package I wrote, but I don't know if it is
> turned on by default. The default might be truncate to zero. I believe
> the denomalized numbers work properly, but they haven't been tested all that
> thoroughly.

I was not sure if it was on by default either.  That's why I put the
x == 0.0 test in.  I believe that your denorm code is used by default,
but I don't think this code is pushing that particular kernel feature.
I need to pull out my IEEE spec to be sure though...

Jon