[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gcl-devel] Re: [Maxima] branch cut fixup for atanh(bfloat(x))

From: Camm Maguire
Subject: [Gcl-devel] Re: [Maxima] branch cut fixup for atanh(bfloat(x))
Date: 01 Mar 2006 15:28:19 -0500
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2


Just a quick update on an older related manner.

In gcl 2.7.0 (unreleased), I've recently found it useful to support
constants (in the 'si package) for +inf, -inf and nan.  These are
actually used as proxies for unbounded real type probing of functions
and type propagation in the compiler to good effect.  In so doing, I
discovered si::*print-nans*, which when set to t makes these
printable.  I've also provided #'si::isfinite.  So far no reading, nor
signed zero, but some convention (as long as it doesn't break ansi)
seems straightforward and even backportable to 2.6 conceivably, as
2.7. is not ready yet.

Take care,

"Robert Dodier" <address@hidden> writes:

> Hi Ray, you wrote:
> > I think there needs to be more discussion about this.  Some Lisps
> > support signed zeroes (cmucl, sbcl), others do not.  Maxima's
> > bigfloats don't support signed-zeroes.
> I don't see how we can have any results which depend on the
> presence of signed zero, be it float, bigfloat, or otherwise.
> On a theoretical level: CLHS doesn't require that CL implementations
> support signed zero. So if we want Maxima to run on a CL implementation
> we can't require any results to depend on the presence of signed zero.
> On a practical level: GCL and Clisp (and doubtless other implementations)
> don't support signed zero. GCL could probably be talked into it.
> Clisp explicitly denies the existence of IEEE 754 special values.
> In any event, for any other implementations signed zero would be
> a portability issue. IEEE 754 can't be considered a standard de facto
> for CL floating point arithmetic.
> I'm not in favor of providing a signed zero for floating point implementations
> which lack it. I think that implies the bigfloat implementation shouldn't
> have one either.
> This isn't the end of the story from my pov but I'll let that be enough for 
> now.
> best,
> Robert
> _______________________________________________
> Maxima mailing list
> address@hidden
> http://www.math.utexas.edu/mailman/listinfo/maxima

Camm Maguire                                            address@hidden
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah

reply via email to

[Prev in Thread] Current Thread [Next in Thread]