[cfp-interest 3523] Re: Rounding Direction Mode or Rounding Direction or Rounding Mode

[Jerome Coonen]

Here is an attempt to tie David's and Jim's replies to Damian's question.

Section 5.3.5.3.3, Characteristics of floating-point types <float.h>, keeps
the historic usage "rounding mode" to refer to rounding direction in
paragraph 23. Jim points out that the more modern usage "rounding
attribute" of 60559 conflicts with the use of "attribute" in C2Y, making
that an unattractive option for C2Y.

The word "constant" Jim used in "constant rounding mode" refers to a static
setting of rounding direction,vs. dynamic setting through
environment function calls. That's a different issue from the usage of
"mode" or "attribute" as the name of the rounding direction.

David pointed out that there are, strictly speaking, other kinds of
"rounding modes", such as the precision level in a wider extended type.
Those cases are arguably less common and less important, making the
enforcement of a specifier for them, like "rounding precision mode", less
onerous. It's not beautiful, but that's yet another matter.

In the absence of user complaints, my conclusion is that C should stick
with the traditional rounding mode usage.

Thanks.

-Jerome Coonen
 650.996.4738
 jcoonen at gmail.com


On Tue, Jun 10, 2025 at 11:07 AM Damian McGuckin <damianm at esi.com.au> wrote:

> On Tue, 10 Jun 2025, Jim Thomas wrote:
>
> > C already uses ?attributes? for something different. See 6.7.13.
> > Considering this, and considering the terminology in IEC 60559 4.2, CFP
> > decided to use "constant (rounding) modes?.
>
> Getting back to the original question, should rounding direction mode
> disappear and rounding direction (when it really means rounding mode) be
> made consistent?
>
> Thanks - Damian
> _______________________________________________
> cfp-interest mailing list
> cfp-interest at oakapple.net
> http://mailman.oakapple.net/mailman/listinfo/cfp-interest
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20250610/6c157e9c/attachment-0001.htm>