[cfp-interest 3415] Re: [SC22WG14.29708] powr(negative, qNaN)

[Jim Thomas]

I too think Joseph is correct about the intention for powr. Here’s a suggestion to make this intention explicit. In F.10.5.7, change:

powr(x, y) returns a NaN and raises the "invalid" floating-point exception for x < 0.

to:

powr(x, y) returns a NaN and raises the "invalid" floating-point exception for x < 0 and all y (including NaN).

The added words follow the style used in F.10.5.8 for rootn(x, 0).


- Jim Thomas



> On Mar 14, 2025, at 5:54 AM, Fred J. Tydeman <tydeman at tybor.com> wrote:
> 
> On Thu, 13 Mar 2025 18:03:24 +0000 (UTC) Joseph Myers wrote:
>> 
>> Do I understand correctly that powr(x, qNaN) should raise "invalid" for 
>> negative x, since such x are outside the domain of the function as defined
>> in IEEE 754?
>> 
>> The relevant statements in Annex F are
>> 
>> "Functions with a NaN argument return a NaN result and raise no
>> floating-point exception, except where explicitly stated otherwise."
>> 
>> and
>> 
>> "powr(x, y) returns a NaN and raises the "invalid" floating-point
>> exception for x < 0."
>> 
>> where the latter doesn't explicitly say "even if y is a NaN", unlike some
>> other cases such as rootn(qNaN, 0), where it's explicit that rootn(x, 0)
>> raises "invalid" for all x including NaN.
>> 
>> IEEE 754 says
>> 
>> "powr(x, y) signals the invalid operation exception for x < 0"
>> 
>> and
>> 
>> "powr(x, qNaN) is qNaN for x  0"
>> 
>> as well as
>> 
>> "Outside its domain an operation shall return a quiet NaN and signal the
>> invalid operation exception.".
>> 
>> If my understanding of the intent is correct then I'll file this as an
>> issue that it should be more explicit that "invalid" is raised in this
>> case of a NaN argument.
> 
> I agree with your understanding.
> 
> 
> 
> ---
> Fred J. Tydeman        Tydeman Consulting
> tydeman at tybor.com      Testing, numerics, programming
> +1 (702) 608-6093      Vice-chair of INCITS/C (ANSI "C")
> Sample C17+FPCE tests: http://www.tybor.com
> Savers sleep well, investors eat well, spenders work forever.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20250314/b9149d33/attachment.htm>