[MPlayer-cvslog] CVS: main/postproc swscale_template.c, 1.116, 1.117

Guillaume POIRIER poirierg at gmail.com
Mon Jun 27 20:27:09 CEST 2005


Hi,

On 6/27/05, Reimar Döffinger <Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
> Hi,
> On Sun, Jun 26, 2005 at 10:53:49PM +0200, Reimar D?ffinger wrote:
> > On Sun, Jun 26, 2005 at 09:33:56PM +0200, Michael Niedermayer wrote:
> > > with gcc yes, otherwise no, either the compiler can create a 64 bit temporary
> > > variable on the stack and use it or it should fail with a fatal error
> > >
> > > but its not as simple as just saying "the developer should put things in
> > > correctly sized variables" because this leads not only to obfuscated and hard
> > > to read source code but also to slow code if the compiler fails to remove the
> > > redundant stuff
> >
> > Hmmm... I wonder, can't we ifdef it to make make gcc use a register on
> > AMD64? I guess there should be enough available (or maybe g constraint
> > would work in all cases??)...
> 
> I was unclear here, I meant g without typecasting to long....

I just tested, it still leads to a crash with GCC-3.3. Is it time for
a GCC bug report?

Guillaume
-- 
A woman isn't complete without a man.
But where do you find a man - a real man - these
days?
  -- Lauren Bacall
My answer: Only at #mplayerdev on the freenode IRC network.




More information about the MPlayer-cvslog mailing list