[Cfp-interest 2767] Re: CFP review of NB comments on CD2 - 20230602

[Hans Boehm]

On Mon, Jun 5, 2023 at 9:07 AM Jim Thomas <jaswthomas at sbcglobal.net> wrote:

>
> We don’t want to say that the floating-point environment must be precise
> at sequence points. That would disallow optimizations like code motion and
> common subexpression elimination that can be safely done between function
> calls.
>
> - Jim Thomas
>
> +1

I think that technically "between" in "between function calls" needs to be
defined in terms of "sequenced before". And making this precise seems both
tricky and unnecessary. (The C++ standard got rid of "sequence points" long
ago because none of us could define precisely enough what that wording
means. I still can't. So I'm generally opposed to adding references to it.)

This text looks to me like it's a normative part of the standard. It
doesn't look to me like it should be. Does the C standard have an easy way
to make it non-normative? Presumably "between function calls" is really
still much too strong, and really only an example? It only needs tobe
"between function calls that might test exception flags"? But I'm not sure
what this kind of negative statement means in normative text.

Hans
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20230605/457a3d9c/attachment-0001.htm>