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

[scott douglass]

On Tue, 2008-06-10 at 03:16, Dennis Handly wrote:
> >From: scott douglass <scott.douglass at arm.com>
> >I'm happy to use 'D<letter>' if there's already a precedent.
> 
> Well, we talked about it for the 3 Dec FP types.
> I just mentioned using it for the char16_t and char32_t.
> 
> >What are the existing 'D<letter>' manglings?
> 
> I think Df, De, Dr.
> The only issue is that D0 D1 D2 are used for destructors so there is
> some lookahead needed.
> 
> >If 'Dh' is available, I change my proposal to:
> ><type> ::= Dh # half-precision floating point (16bit)
> 
> Yes, Dh is free.
> 
> >From: Mark Mitchell <mark at codesourcery.com>
> >I don't see any precedent there -- but I think we could do that.
> 
> Yes, using D, seems to give us 26*2 types.
> 
> >Here is another possible trick that we could use for both DFP and 
> >half-precision float...  Since we have:
> >So, we could use "uh" for half-precision float.  That's the same number 
> >of characters and avoids using any new encoding space?
> Mark Mitchell
> 
> It seems kind of hokey.  Almost as bad as my D[0-9] vs D[a-zA-Z] proposal.
> And we can include those two char types.

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.  Or do you think the DFP types are going to switch to 'u<letter>'?


-- 
IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium.  Thank you.