[FFmpeg-devel] [RFC] clobbers for XMM registers

Michael Niedermayer michaelni
Tue Sep 28 14:21:17 CEST 2010


On Tue, Sep 28, 2010 at 08:00:28AM +0100, M?ns Rullg?rd wrote:
> Michael Niedermayer <michaelni at gmx.at> writes:
> 
> > On Mon, Sep 27, 2010 at 09:15:24PM -0400, Ronald S. Bultje wrote:
> >> Hi,
> >> 
> >> 2010/9/27 M?ns Rullg?rd <mans at mansr.com>:
> >> > Michael Niedermayer <michaelni at gmx.at> writes:
> >> >> On Tue, Sep 28, 2010 at 01:51:33AM +0200, Michael Niedermayer wrote:
> >> >>> On Sat, May 29, 2010 at 08:27:58PM +0200, Reimar D?ffinger wrote:
> >> >>> > Hello,
> >> >>> > this is not suitable by far (among other things, ugly parts, split asm code
> >> >>> > where a clobber is likely to introduce bugs, only tested to compile,
> >> >>>
> >> >>> > did not find a nice "dummy" clobber constraint - particularly important since
> >> >>>
> >> >>> try "%esp" ;)
> >> >>>
> >> >>>
> >> >>> > I suspect older gccs won't accept %xmm clobbers, ...).
> >> >>>
> >> >>> the oldest i have here is 3.3 and it supports it
> >> >>
> >> >> has someone tested xmm clobbers on fate alraedy?
> >> >> if not i think we should commit one and see what fails
> >> >> then we can decide if we need a macro wraping them at all
> >> >
> >> > We tested. ?It failed.
> >> 
> >> See r24931 and r24934 in vp3dsp_sse2.c.
> >> 
> >> http://archives.free.net.ph/message/20100825.195706.b4de9543.en.html
> >> http://archives.free.net.ph/message/20100825.203635.00f52a36.en.html
> >
> > do you happen to remember the error messages and which compiler version
> > was affected?
> 
> All BSD and all Windows.  See for example
> http://fate.ffmpeg.org/x86_32-dragonflybsd-gcc-4.1.2/20100825202620
> http://fate.ffmpeg.org/i686-pc-mingw32-gcc-4.2/20100825201526
> 
> libavcodec/x86/vp3dsp_sse2.c:165: error: unknown register name '%xmm7' in 'asm'

could be a build system bug, aka missing -msse

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Awnsering whenever a program halts or runs forever is
On a turing machine, in general impossible (turings halting problem).
On any real computer, always possible as a real computer has a finite number
of states N, and will either halt in less than N cycles or never halt.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100928/228b6bb9/attachment.pgp>



More information about the ffmpeg-devel mailing list