Re: +-Inf and NaN

On Thu, Feb 22, 2024 at 5:01 PM James Cloos <[email protected]> wrote:

> >>>>> "SM" == Stavros Macrakis <[email protected]> writes:
>
> SM> Lisp arithmetic should be compatible with IEEE float arithmetic.
>
> There are cases where other forms of floats make sense.
>
> Such as posits or arbitrary precision floats.
>
> (The underlying hardware may lack fp, or may use posits or the like.)
>
> Packages like maxima should work even when non-754 formats are used.
>

Is there any existing lisp that doesn't support IEEE single and double
precision floats?  I'm not aware of any.  I don't see any chip maker
supporting posits and removing IEEE floats anytime soon.  That might break
all kinds of important code.  (I saw Kahan debate Gustafson about this a
few years ago at an ARITH conf.  Really interesting.)

Clisp has long-floats that are arbitrary precision.  Cmucl has
double-double-floats.  I think maxima might still support building with
these as the default float type, but since they're significantly slower
than IEEE floats, I don't think anyone uses this ability.  (And if the code
still exists, I'd propose removing it, at least for cmucl.)


-- 
Ray

_______________________________________________
Maxima-discuss mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/maxima-discuss