[Cfp-interest 2118] Re: AI to update proposal about C overflow and underflow definitions

[Jim Thomas]

> On Aug 22, 2021, at 8:41 PM, Damian McGuckin <damianm at esi.com.au> wrote:
> 
> 
> On Sun, 22 Aug 2021, Jim Thomas wrote:
> 
>> Action item:
>> 
>>          Jim: Update N2746 with CFP 2090.
>> Here?s as proposal update:
>> https://wiki.edg.com/pub/CFP/WebHome/C23_proposal_-_overflow_and_underflow_
>> definitions-update-20210822.pdf
> 
> Does the first sentence imply 2 cases, i.e.
> 
> The result underflows if a nonzero result value with ordinary accuracy would have magnitude (absolute value) less than the minimum normalized number in the type, or a zero result that is not an exact zero 249)...

Not quite. An intention is to avoid requiring the implementation (which need not be correctly rounded or follow Annex F) to detect exact zeros. If the implementation follows Annex F (or specifies zero cases) then a specified zero case does not underflow. If an implementation delivers a zero result where no exact zero has been specified, then the result underflows. 

A fair question is whether the latitude in the definition is unnecessary, because it’s easy enough to determine exact zeros for the particular functions in the C library. If so, we could just say “however, an exact zero result does not underflow.”

- Jim Thomas

> 
> Stay safe - Damian
> 
> Pacific Engineering Systems International, 277-279 Broadway, Glebe NSW 2037
> Ph:+61-2-8571-0847 .. Fx:+61-2-9692-9623 | unsolicited email not wanted here
> Views & opinions here are mine and not those of any past or present employer