[Cfp-interest] Fwd: (SC22WG14.12927) RE: RE: a term for float, double, and long double

[James W Thomas]

Begin forwarded message:

> From: James W Thomas <jaswthomas at sbcglobal.net>
> Subject: Re: (SC22WG14.12927) RE: RE: a term for float, double, and long double
> Date: May 20, 2013 9:03:30 AM PDT
> To: <beh at ix.netcom.com>
> Cc: "'SC22 WG14'" <sc22wg14 at open-std.org>
> 
> C99 changed the type structure
> 
> 	signed integer types
> 
> to
> 
> 	signed integer types
> 		standard signed integer types (adding long long)
> 		extended signed integer types
> 
> TS 18661 Part 2 changes
> 
> 	real floating types
> 
> to
> 
> 	real floating types
> 		generic/traditional floating types
> 		decimal floating types
> 
> (Part 3 tucks another collection of types under real floating types.)
> 
> So, we are following the model for integer types in the C Standard, except for the names "standard" and "extended". 
> 
> The term "extended format" has a well-established and very different meaning in the IEC 60559 FP Standard. Most of the new types we're proposing are not "extended" in the IEC 60559 sense. Calling all of them "extended" would cause confusion.
> 
> Referring to float, double, and long double as the "standard floating types" might be reasonable. Part 2 proposes changes to the C Standard that make the decimal floating types conditionally normative, not just possible extensions (like the extended integer types). Defining "standard floating types" to exclude conditionally normative floating types doesn't seem like a clear win. Is "standard" a better term than "traditional"?
> 
> -Jim
> 
> On May 19, 2013, at 7:53 AM, "BEH Netcom" <beh at ix.netcom.com> wrote:
> 
>> Then I favor some clarity here, because different terms are being introduced to mean essentially the same thing as terms already defined by the Standard.
>>  
>> 1. The term 'real floating types' should always mean float, double and long double.  That's already defined.  I am not in favor of starting to call them something else, regardless of what the something else is.
>>  
>> 2. The term 'generic floating type' should not be used at all. We already have a keyword _Generic, several 'generic functions', as well as type-generic math that applies to floating point types.  It's simply too easy to confuse the terms 'type-generic' with 'generic type'.
>>  
>> If the intent is, as stated below, to add types to those defined as 'real floating types' (float, double, long double) then I suggest an approach modeled on that used in 6.2.5, Types, for 'standard integer types'.  Additional integer types are allowed, they are implementation-defined, and they are called 'extended integer types'.
>>  
>> It seems to me that model works for the additional types to be added to 'real floating types'. They can be called 'extended real floating types'.  The difference is they are defined in the specification rather than being implementation-defined, but that should not matter.  I believe that model is clear and consistent.
>>  
>> Barry
>>  
>>  
>> From: owner-sc22wg14 at open-std.org [mailto:owner-sc22wg14 at open-std.org] On Behalf Of James W Thomas
>> Sent: Saturday, May 18, 2013 5:00 PM
>> To: beh at ix.netcom.com
>> Cc: 'SC22 WG14'
>> Subject: (SC22WG14.12926) RE: a term for float, double, and long double
>>  
>> We need the specification in the Standard for real floating types to apply to the new types we're adding in TS 18661 Parts 2 and 3. So, "real floating types" need to include more than float, double, and long double. It has been convenient to have a collective term to refer to float, double, and long double, and only those types. You can search through Part 2 for "generic floating type"  to see how much it is used there. 
>>  
>> -Jim
>>  
>> On May 17, 2013, at 1:55 PM, BEH Netcom <beh at ix.netcom.com> wrote:
>>  
>> 
>> The C Standard already has a term for these types:
>>  
>> 6.2.5;p10
>> There are three real floating types, designated as float, double, and long double.
>>  
>> Is there a real need to come up with yet another term for these three types?
>>  
>> Barry
>>  
>> From: owner-sc22wg14 at open-std.org [mailto:owner-sc22wg14 at open-std.org] On Behalf Of James W Thomas
>> Sent: Friday, May 17, 2013 9:57 AM
>> To: SC22 WG14
>> Subject: (SC22WG14.12924) a term for float, double, and long double
>>  
>> At Delft, in the meeting segment on TS 18661 Part 3, concern was expressed about the use of the term "generic floating types", which is introduced in Parts 2 and 3 (and the existing Decimal TR), to refer to the float, double, and long double types (and not to the other floating types introduced in those specifications). The issue is the possible confusion with other uses of the term "generic" in the C Standard. At the CFP teleconference yesterday, we considered other terms and decided to recommend "traditional". For example, Part 2 would say: 
>>  
>>           There are three traditional floating types, designated as float, double, and long double.
>>  
>> and:
>>  
>>           Together, the traditional floating types and the decimal floating types comprise the real floating types.
>>  
>> Other terms we considered include "general", "default", and "classical".
>>  
>> If you have concerns with this use of the term "traditional", please respond by May 31.
>>  
>>  
>> - Jim Thomas
>>  
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20130520/3d0829a4/attachment.html