[PATCH] Fix AAPCS-VFP non-compliance when returning HFA from variadic function

Renato Golin renato.golin at linaro.org
Thu Jan 16 09:21:32 PST 2014


On 16 January 2014 16:02, Oliver Stannard <oliver.stannard at arm.com> wrote:

> This seems like a fairly fragile way to do things. Does anyone know what
> the rationale behind doing this in clang is?
>

AFAICR, Clang was working around back-end deficiencies on many
procedure-call decisions, which didn't get fixed because Clang was working
around it. When I got to connect our front-end I ended up repeating the
same mistakes. I've been told that that's llvm-gcc's fault, but that was
too long ago...

The main problem now is that changing procedure-call functions now would
need some bigger coordination than just us ARM folks, and could prove to be
quite difficult.

That being said, I'm always up for a good change. ;)

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140116/4efbbf7c/attachment.html>


More information about the cfe-commits mailing list