[Cfp-interest 2188] Re: AI about new wording for unaccepted change

[Mike Cowlishaw]

If that's the case, maybe those uses should be revisited.
 
But if the meaning of equivalent is "one expression can replace the other", then it definitely should not be used here, because 1.5 and 1.50 certainly cannot be interchanged.
 
Mike


  _____  

From: Jim Thomas [mailto:jaswthomas at sbcglobal.net] 
Sent: 02 October 2021 22:21
To: Mike Cowlishaw
Cc: CFP
Subject: Re: [Cfp-interest 2171] AI about new wording for unaccepted change



On Oct 1, 2021, at 5:54 AM, Mike Cowlishaw <mfc at speleotrove.com> wrote:

Depending on how 'value' is used elsewhere, I prefer (b) because I think many people consider 'different values' to be synonymous with 'do not compare equal' .. so (a) appears to be an oxymoron.
 
But (b) has the same problem, in truth, because the primary sense of 'equivalent' is 'equal in value'.


The phrase “are not equivalent" is used six time in the same subclause. The subclause is about expression transformations (optimizations). The term “equivalent” is used to mean one expression can replace the other, without any detectable behavior change. Also, the term “equivalent” is used in other places in the C standard, with a broader meaning than “equal in value”, for example in saying code sequences are equivalent. 

- Jim Thomas



 
Perhaps:
 
  The results are not the same because they have different quantum exponents. 

 
Mike


  _____  

From: Cfp-interest [mailto:cfp-interest-bounces at oakapple.net] On Behalf Of Jim Thomas
Sent: 30 September 2021 23:47
To: CFP
Subject: [Cfp-interest 2171] AI about new wording for unaccepted change


Action item: 


    Jim: Propose new wording for N2716's change that WG14 did not accept as is.


The suggested change in question is:


Page 450, paragraph 4, in the Example, change:


The results are numerically equal, but have different quantum exponents, hence have different values.

to

The results are equal, but have different quantum exponents, hence have different values.


The immediately preceding text is:


3 For expressions of decimal floating types, transformations must preserve quantum exponents, as well as numerical values (5.2.4.2.3). 

4 EXAMPLE 1. × x → x is valid for decimal floating-point expressions x, but 1.0 × x → x is not: 
1. × 12.34 = (+1, 1, 0) × (+1, 1234, −2) = (+1, 1234, −2) = 12.34 
1.0 × 12.34 = (+1, 10, −1) × (+1, 1234, −2) = (+1, 12340, −3) = 12.340


A minimal change that responds to the reported WG14 discussion [Cfp-interest 2128] would be:


a) The results compare equal, but have different quantum exponents, hence have different values.


However, we haven’t used “compare equal” in the subclauses. I think a more direct statement would be:


b) The results are not equivalent because they have different quantum exponents.



Please let me know ASAP if you prefer a) or b) or something else. 

- Jim Thomas







-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20211003/a95e8024/attachment-0001.htm>