Re: Is it possible to support vfp abi for wince 6.0?
Hi Andy,
Andy Wu escreveu:
> Hi Pedro,
>
>
> On 10/24/06, *Pedro Alves* < pedro_alves@...
> <mailto:pedro_alves@...>> wrote:
>
> Hi Andy,
>
> Andy Wu wrote:
>
> > Hi all,
> >
> > The wince 6.0 will be released soon, which will update the ABI to
> > support VFP (ARM11). Is it possible to enable the hard floating
> point
> > ABI to cegcc? Or anybody know any floating support in the current
> > cegcc code. What I know is that the current cegcc only support
> > soft-abi (hardware floating VFP code can be generated in the
> function
> > but the interface is still based on soft-ABI)...
> >
>
> Could you please share the pointer to the source of this
> information? I
> can't find anything related to an ABI change.
> I would be very interested.
>
> The only link I have regarding VFP on WinCE is this:
>
http://msdn.microsoft.com/chats/transcripts/mobileembedded/06_0830_msdn_wince.aspx
> What I understand from that discussion is that an OEM can
implement an
> optimized floating point library using VFP
> if the target has support for it, but it doesn't mention anything
> about
> the ABI being incompatible.
>
>
> I don't think microsoft will change any ABI spec based on ARM's spec.
> Here is the information from ARM:
> http://www.arm.com/products/DevTools/ABI.html
>
I don't think so too, that's why I asked for pointers. :)
> I'm not sure if gcc support vfp ABI at elf or other obj format... If
> the answer is yes, I think enable the vfp ABI for
> wince pe is not ver difficult. I think this is a very cool and new
> feature for this project. It will be very useful when
> wince 6.0 released.
>
> Do you any ideas on this topic?
It doesn't have to do with the object format. It has to do with
gcc's ability to emit vfp instructions and still be compatible with
the float ABI required by WinCE.
(soft abi stores floats/doubles always in big endian format, and
softvp/vfp uses the native endianess.)
Gcc doesn't emit vfp instructions when using
-mfloat-abi=soft, only when using `-mfloat-abi=softfp -mfpu=vfp`
Since we must stay compatible with WinCE's float ABI, we can't
change to softfp ABI, like linux arm distributions are doing.
To be able to emit vfp instructions directly, and still be compatible
with WinCE, someone will have to change gcc. A very nice project indeed.
(Note that the assembler (gas) accepts VFP, it is gcc that doesn't emit it.)
The float ABI is something that has been on my todo list,
like confirming the alignments, packing, endianess, calling conventions
related to floats/doubles.
I would really like to find some authorative (MSDN/Microsoft)
information about this.
Cheers,
Pedro Alves
P.S.: Please don't top post, inline/bottom post instead.
-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
RSS Feed