C++ Logo


Advanced search

Re: [wg14/wg21 liaison] Comments on P2569 *_HAS_SUBNORM==0 implies what?

From: Jens Maurer <Jens.Maurer_at_[hidden]>
Date: Wed, 30 Mar 2022 20:56:11 +0200
On 30/03/2022 19.46, Fred J. Tydeman via Liaison wrote:
> On Mon, 21 Mar 2022 16:05:56 +0000 Jonathan Wakely via Liaison wrote:
>> The new denorm_min introduced by P1841 is supposed to make it unnecessary,
>> but doesn't really do so, which is the subject of question 1 in P2551.
>> Question 4 in P2551 is also related.
> When will those questions be answered?

They were answered in the LEWG session on Tuesday.
The minutes aren't available yet, it seems, but my recollection is:

Q1: We consider the existence of a corresponding bit-pattern
decisive for any traits asking about special kinds of numbers
(denormals, negative zero, infinity), regardless of whether
those could be produced by arithmetic operations (maybe
because the hardware can't do it or because the CPU is in a
mode where those aren't produced).

Q2: Disable traits for bool.

Q3: Disable floating-point traits for integers.

Q4 + Q5: no idea

There was a concern that this would make the existing
facility std::numeric_limits deviate from the new traits
(or we'd need to fix the definition of std::numeric_limits),
but people seemed generally ok with that.

> Will they be answered by a C++ numerics sub-group,
> or the entire committee?

SG6, the C++ numerics sub-group, did discuss these issues, too.

> Based upon other emails here, C++ does not care if C
> were to remove the *_HAS_SUBNORM macros.
> Correct?

That's my understanding.


Received on 2022-03-30 18:56:30