[Cfp-interest 1829] Re: Preferred quantum exponent

[Jim Thomas]

> On Oct 26, 2020, at 10:52 PM, Fred J. Tydeman <tydeman at tybor.com> wrote:
> 
> I believe that the following (or similar):
> 
>  The preferred quantum exponent for finite specified return values in this sub-clause is 0.
> 
> should be added to C23 Annex F in F.10.
> 
> Reason
> 
> 5.2.4.2.3#7 has a table of preferred quantum exponents.
> For example, cbrt has Q(x)/3 as its preferred quantum exponent.
> 
> Annex F has many specifications of return values.
> For example, cbrt(+/-0) return +/-0.
> 
> So, what is supposed to be returned for cbrtd64(0.e-300DD)?
> Is it 0.e-100DD based upon the table in 5.2.4.2.3?

Yes. This is what is specified.

> Or, should it be 0.e0DD (the natural zero)?

How is this different from 0e-1 * 0e-1 = 0e-2?

- Jim Thomas

> 
> I raised a similar question to the IEEE-754 email list and the one
> reply I got is:  This sounds like a bug fix for the next revision -
> 
> 
> ---
> Fred J. Tydeman        Tydeman Consulting
> tydeman at tybor.com      Testing, numerics, programming
> +1 (702) 608-6093      Vice-chair of PL22.11 (ANSI "C")
> Sample C99+FPCE tests: http://www.tybor.com
> Savers sleep well, investors eat well, spenders work forever.
> 
> _______________________________________________
> Cfp-interest mailing list
> Cfp-interest at oakapple.net
> http://mailman.oakapple.net/mailman/listinfo/cfp-interest