On 2020-11-13 20:02 -0500, John Cowan wrote:
> I've added an optional argument "float" to make-nan, which if present says
> that the NaN has the same number of bits of representation as the float.
> This way we can produce, say, binary32 NaNs if and only if the Scheme
> implementation has binary32: you call make-nan with a fourth argument of
> 0.0f0.

How can we meaningfully implement this in the C sample implementation?
Currently we just have two sets of procedures, one for doubles and one
for floats.

--
Wolfgang Corcoran-Mathe  <xxxxxx@sigwinch.xyz>

"Invent and fit; have fits and reinvent!" --Alan J. Perlis