Re: largest finite float

At 10:15 AM -0500 1/24/02, zongaro@ca.ibm.com wrote:

>      As long as we're on the topic of subnormalized numbers, it's 
>not clear to me whether they were intended to be part of the value 
>space.  The description of the value space I've quoted above admits 
>the subnormalized values to the value space - they are those values 
>for which e=-149 and 0 < m < 2**23.  However, literals in the 
>lexical space map to the closest *normalized* value.  That would 
>mean that there are values in the value space to which no value in 
>the lexical space will map.  Any revision of the description of 
>float needs to answer this question.

This sounds like an erratum candidate to me.  It's my understanding that
subnormalized numbers are included, and the rounding algorithm maps to the
closest value in the value space, be it normalized or subnormalized.

I suspect any other interpretation would have the entire community of
float/double users up in arms.

I believe the question of whether extremely large numbers should "round"
to the largest representable or to +infinity was discussed and resolved
in favor of +infinity, with the cutoff as I described.  I don't recall
this one (subnormals) being discussed.
-- 
Dave Peterson
SGMLWorks!

davep@acm.org