[Cfp-interest 3355] Re: Fwd: [isocpp-wg14/wg21-liaison] SG22 telco : P3477R0 and P2746R0

[Hans Boehm]

>From my perspective, for P2746, yes.

The expectation is that floating point rounding modes will end up as a
point of divergence between the two languages, but it's unclear how to
avoid that. The C23 static rounding mode pragma is not likely to be
accepted into C++. The current C++ library proposal relies on member
functions, and is not C friendly, though that could possibly be
addressed. The IEEE standard seems to target a scope-based mechanism
for setting rounding modes, which some of us dislike. Thus the CFP
group would have to stretch its charter to accept a library mechanism,
along the lines of P2746.

On Thu, Jan 9, 2025 at 3:07 AM Nina Dinka Ranns
<dinka.ranns at googlemail.com> wrote:
>
> Hello,
>
> has this been appropriately discussed ?
>
> If so, I would like to schedule an SG22 meeting for it.
>
> Thank you,
> Nina
>
> On Tue, 7 Jan 2025 at 16:13, Hans Boehm <boehm at acm.org> wrote:
> >
> > Thank you! Fixed in the next revision.
> >
> > Hans
> >
> > On Sun, Jan 5, 2025 at 8:39 AM Damian McGuckin <damianm at esi.com.au> wrote:
> > >
> > > On Sat, 4 Jan 2025, Damian McGuckin wrote:
> > >
> > > > Unlike R0, it is in 2024, not 2023. So
> > > >
> > > >       https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2024/p2746r6.pdf
> > >
> > > Unless I am opening the wrong file, .....
> > >
> > > In that document, the link in the PDF to apparently
> > >
> > >         High Perforance Correctly Rounded Math LIbraries
> > >         for 32-bit Floating Point Representations
> > >
> > > by Lim and Nagarette, which is quoted as
> > >
> > >         https://dl.acm.org/doi/abs/10.1145/3385412.3386004
> > >
> > > resolves to:
> > >
> > >         Debugging and detecting numerical errors in computation with posits
> > >
> > > (by CHowdhray and those Authors).
> > >
> > > The correct link should be
> > >
> > >         https://dl.acm.org/doi/pdf/10.1145/3453483.3454049
> > >
> > > Thanks - Damian