Ah, yes, I understand. I was not attempting to voice an opinion on this. I only wanted to draw attention to newer floating-point formats that may be worth considering in the design of a new type trait.

Some floating-point arithmetic behavior may be chosen at runtime, e.g., whether subnormal results are flushed to zero. So, some behavior is not easy to classify.



-----Original message-----
From: Sebastian Wittmeier via Std-Proposals
Sent: April 18, 2026, 23:12 (UTC)
To: std-proposals_at_[hidden]
Cc: Sebastian Wittmeier
Subject: Re: [std-proposals] ISO/IEC 60559 Classification for std::numeric_limits
It is

Jan as OP suggested

 
"We ought to have some portable way to classify a floating-point type according to its ISO/IEC 60559 format."

 
Should only the format be part of the classification or the precision or the behavior, too.

 
Or should the classification just give the notion that it is an ISO 'compatible' format or an implementation defined format?


 
-----Ursprüngliche Nachricht-----
Von: Paul Caprioli <paul_at_[hidden]>
Gesendet: Sa 18.04.2026 23:04
Betreff: RE: [std-proposals] ISO/IEC 60559 Classification for std::numeric_limits
An: std-proposals_at_[hidden];
CC: Sebastian Wittmeier <wittmeier_at_[hidden]>;
And is it only about the format or also about the behavior of the types under common mathematical operations?
And format (as I think was mentioned) can be storage format or range of mantissa and exponent.

 
I don't know what "It" means. The spec does define conversion from wider floating point types and describes saturating and non-saturating modes.
So, one could perform common mathematical operations in infinite precision and then narrow the result to quarter precision. However, I don't see that described.

The spec "provides two binary interchange formats for floating point encodings."

--
Std-Proposals mailing list
Std-Proposals_at_[hidden]
https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals