[cxx-abi-dev] IEEE 754r half-precision fp mangling?

[Dennis Handly]

>From: scott douglass <scott.douglass at arm.com>
>If 'Df', 'De' and 'Dr' are already (nearly?) in use, I think 'Dh' is the
>more reasonable choice.  Unless there are objections, we'll start using
>that.

Ok.

>From: Mark Mitchell <mark at codesourcery.com>
>No, let's just go with Dh; that seems simplest given where we are. 

Similar for those char16_t & char32_t ones?

>From: scott douglass <scott.douglass at arm.com>
>If I knew what the DFP sizes were, I'd put them in as well.

Do we have any rationale or foot notes why we chose "D"?
And for demanglers to caution in that D0 ... D2 is used for function names?

>From: Mark Mitchell <mark at codesourcery.com>
>DFP folks: would one of you please fill in proper names for the DFP 
>mangling types?

>From: PremAnand M Rao <premanand.rao at hp.com>
>> +              ::= De # IEEE 754r decimal floating point (size?)
>> +              ::= Df # IEEE 754r decimal floating point (size?)
>> +              ::= Dh # IEEE 754r half-precision floating point (16 bits)
>> +              ::= Dr # IEEE 754r decimal floating point (size?)

>I think there is an error in the above, there should not be both a "De" 
>and a "Dr".  I believe Michael suggested "Dr" for 128-bits decimal 
>float, but Dennis and Daveed expressed a preference for "De" (instead of 
>"Dr") to match the binary floating point mangling.

I just copied Dr from a previous (wrong?) message.

>I believe the above table should be corrected with:
>Df # IEEE 754r decimal floating point (32 bits)
>Dd # IEEE 754r decimal floating point (64 bits)
>De # IEEE 754r decimal floating point (128 bits)
PremAnand.

And their C names are: _Decimal32, _Decimal64, _Decimal128, 

>From: PremAnand M Rao <premanand.rao at hp.com>
>And I would also like to ask again the opinion of all on the list if 
>"Dh" really is appropriate for the half-precision floating point types. 

We are using "D" as an extension letter, it doesn't really mean "decimal"
anymore.  :-)

>half-precision floating point type is a binary floating point type whereas
>the other "D[fde]" types are decimal floating point types.
PremAnand.

That's correct but we have recently gotten requests for 6 new types and
rather than waste single letters, my suggestion was just use "D" for
a "family" of 2*26 more builtin types.