[ANSI-Smalltalk] Agenda items...

[Andres Valloud]

Paolo,

This is why I think we should separate the issue of numerical quantity from
the issue of numerical representation.

There is a double the value of which is the integer 2^1023 exactly.

There is no double the value of which is the integer 2^1024 exactly.

How we choose to represent such numbers in the context of different computer
languages is a different issue.

Now, we can easily add emaxLIA, emaxIEEE and
emaxAnotherRepresentationStrategy.  However, the only thing that remains
stable, independently of how we arbitrarily choose to represent doubles in
source code, is the implementation limit of the binary representation ---
which is codified as a standard and is something that every source code
representation scheme will have to abide by.  Therefore, I think emax should
refer to the characteristic behavior of the binary representation of
doubles.

Thanks,
Andres.

On Dec 3, 2007 7:09 AM, Paolo Bonzini <paolo.bonzini at lu.unisi.ch> wrote:

> > Note that this is so regardless of whether we
> > choose to write mantissas in LIA or in IEEE, because we still agree that
> > 2^1023 is a double while 2^1024 is not.
>
> The point is that neither 2^1023 nor 2^1024 are doubles, because the
> mantissa is 1 (out of range!).  0.5*2^1024 is a double, and 0.5*2^1025
> is not a double.  That's why *in C* emax is 1024 and Smalltalk should do
> the same.
>
> Paolo
>
> _______________________________________________
> ANSI-Smalltalk mailing list
> ANSI-Smalltalk at lists.openskills.org
> http://lists.openskills.org/cgi-bin/mailman/listinfo/ansi-smalltalk
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.openskills.org/pipermail/ansi-smalltalk/attachments/20071203/4211be0a/attachment.html