Re: [Stds-754] nomenclature for representable entities

[Jim Thomas <jwthomas@xxxxxxxxxx>]

Based on a quick look at all uses of "number" in the draft, it appears that most are intended to exclude NaNs, which is consistent with Warren's suggestion.  For example, 7.3.1 says "if one operand is a number and the other a NaN", 7.5.1 refers to "treating numbers and NaNs alike", 7.10 has "totalOrder(number, +NaN)" implying that the first operand is not a NaN.  I think following Warren's suggestion will require fewer and less awkward changes.

-Jim

"Ferguson, Warren E" wrote:

At the last style review we were asked to come up with a taxonomy of "representable entities". Here's my response:

The "representable entities" include NaNs, so by double negation anything not NaN must be "a number".

Numbers fall into two classes: "infinity" and "not infinity = finite".

Finites fall into two classes: "zero" and "nonzero".

I guess nonzero includes "normal" and "unnormal" - but I'm not sure if this is as useful as it used to be for just 754 binary numbers.

I don't want to restrict unnormal numbers to 754's denormal entities because there a denormal had a reserved exponent.

Proposal: zero + nonzero + infinite + NaN = representable entitites.

Warren Ferguson

_______________________________________________
Stds-754 mailing list
Stds-754@xxxxxxxxxxxx
http://mailman.oakapple.net/mailman/listinfo/stds-754