[Cfp-interest 2622] Re: CFP position on US 5-018?

[Jim Thomas]

CFP agreed to US 5-018 at its 2023/01/05 continuation meeting. [Cfp-interest 2590] Preliminary WG14 IEEE 754-C binding meeting minutes - 2023/01/05 says "US5-18: No issues. - Agreed."

The macros as introduced by N2379 allowed a value to indicate the type supported both IEC 60559 values and IEC 60559 operations. A recent change recommended by CFP removed this macro value because there was no indication of what the binding for those operations might be since Annex F conformance was not required. Now the macros attempt to characterize the values in the type without referring to the arithmetic. Implementations with IEC 60559 formats generally produce normal result values according to IEC 60559, even if they don’t support IEC 60559 in other regards. But to have the macros say an IEC 60559 -0 value is included didn’t seem helpful without knowledge that the arithmetic would treat the value like -0 in IEC 60559. Similarly for subnormal numbers, infinities and NaNs. 

- Jim Thomas

> On Jan 19, 2023, at 2:41 PM, Hubert Tong <hubert.reinterpretcast at gmail.com> wrote:
> 
> Hello,
> 
> I do not believe that the proposed resolution provided in N3067 for US 5-018 necessarily reflects the intent of the committee because precision and exponent range is not sufficient to reflect the availability of signed zeros, signed infinities, etc.
> 
> Has the CFP position on this been established? Where would I find this information?
> 
> Thanks,
> 
> 
> Hubert Tong