[Cfp-interest 2013] Re: AIs related to range errors clarifications

[Jim Thomas]

> On May 28, 2021, at 4:47 PM, David Hough CFP <pcfp at oakapple.net> wrote:
> 
> Thinking of the infinitely-precise result I thought
> 
> "or if it is inexact and has magnitude 
> no greater than 
> the minimum normalized number in the type"
> 
> could be
> 
> "or if it is inexact and has magnitude 
> less than 
> the minimum normalized number in the type"
> 
> But if we are talking about the rounded result, it could be inexact
> and equal to the minimum normal, which is the confusing case for most
> people.
> 

In the wording I suggested in email yesterday:

> The result underflows if it is nonzero and has magnitude (absolute value) less than the minimum normalized number in the type, or if it is inexact and has magnitude no greater than the minimum normalized number in the type.

the result is the final result.

On further thought, I don’t think my suggestion above is better than what CFP agreed to send to WG 14:

The result underflows if the a nonzero result value with ordinary accuracy would have magnitude (absolute value) no larger than the minimum normalized number in the type.

so will follow up on the action item to submit it. If CFP finds better wording we can submit an update proposal. This reduces our risk of not getting the proposal in before a WG 14 deadline for new proposals.

- Jim Thomas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20210529/6048a589/attachment.htm>