[Cfp-interest 2001] Re: AI about IEC 60559 binding

[Jim Thomas]

Hi Paul,

The math functions as specified in F3/F10 are not the IEC 60559 recommended mathematical operations. I think the new text makes that clear. The F3/F10 math functions do correspond to the IECC 60559 operations in the sense of providing approximations to the same mathematical functions and handing special cases the same way (to the extent possible without correct rounding). An implementation can provide the IEC 60559 operations as C functions with the cr_ prefix.

This was a design decision made in consideration of implementation practicality and general user needs (as seen by CFP).

- Jim Thomas

> On May 12, 2021, at 11:04 PM, Paul Zimmermann <Paul.Zimmermann at inria.fr> wrote:
> 
>       Hi Jim,
> 
> there is a small issue here. IEEE 754 (and thus IEC 60559 I believe)
> recommends additional mathematical functions like exp. But if these functions
> are present, they *should* be correctly rounded, and I believe this holds
> whatever the rounding mode. Thus when we write "[12] Whether the functions
> ... honor the rounding direction mode", this cannot apply to the functions
> mentioned in Section 9 of IEEE 754-2019.
> 
> All current mathematical libraries are not correctly rounded, thus their
> mathematical functions like exp have nothing to do with Section 9 of
> IEEE 754-2019 (they should be called non_standard_exp).
> 
> Does the F.3/F.10 text correspond to the IEEE 754/IEC 60559 functions,
> or to the functions provided by the current mathematical libraries
> (which have nothing to do with IEEE 754/IEC 60559)?
> 
> Sorry to provide more questions than solutions.
> 
> Best regards,
> Paul
> 
>> From: Jim Thomas <jaswthomas at sbcglobal.net>
>> Date: Wed, 12 May 2021 15:05:59 -0700
>> 
>> Action item:
>> 
>>    Jim: Propose CFP 1952 with the addition of a title for the second
>> table
>> to WG14.
>> 
>> Below is a draft proposal. I didn’t add a title for the second table,
>> because I couldn’t think of a good one, and given the rest of the changes it
>> didn’t seem needed. If we decide otherwise, and come up with a title, it can
>> easily be added. 
>> 
>> https://wiki.edg.com/pub/CFP/WebHome/C23_proposal_-_IEC_60559_bindings-20210512.pdf

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20210513/0e4d28a2/attachment.htm>