[Gcl-devel] RE: IEEE floating point in GCL

[Stavros Macrakis]

> No onle is working on this to my knowledge, though it
> would be straightforward to export some C facility like isnan,
> isfinite, etc.  We use these already in num_co.c.

Camm--

Thanks for the answer.

I don't think it would be enough to export isnan, etc., since GCL
doesn't let you generate NaNs, negative zero, and infinity.  GCL is
doing the right thing here, since all existing Lisp code expects floats
NOT to have these things.

To really support IEEE float, you'd need to support extended precision,
rounding modes, etc.  The rounding modes could be done in the usual way,
namely as a global setting, when they are used for testing (checking
rounding sensitivity of code), but not when they're used for interval
arithmetic....

What's worse, it's hard to get access to these facilities even through
C.  And I don't think there are standard language bindings in Common
Lisp for these facilities....

So may be best to do nothing for now.  When UFFI comes, we can fool
around with various loadable routines.

I probably shouldn't even have brought the topic up, but I was curious
to know if anyone had been thinking about it.

      -s