[Ffmpeg-devel] gcc 4x compilation issues on ppc

[Diego Biurrun]

On Sat, Mar 19, 2005 at 11:06:02PM +0100, Michael Niedermayer wrote:
> 
> On Saturday 19 March 2005 19:12, Bret Hughes wrote:
> > On Sat, 2005-03-19 at 06:25, James A. Morrison wrote:
> > > Bret Hughes <bhughes at elevating.com> writes:
> > > > vu32 has this definition:
> > > > #define vu32(v) ((vector unsigned int)(v))
> > > >
> > > > I can't tell what is happening here. mzero is defined as a float so
> > > > somehow these statements are assigning a float value to mzero.  is this
> > > > a cast and would casting the righthand side of the assignment to a
> > > > float be correct?
> > > >
> > > >   mzero = ((vector float)vec_splat_u32(-1));
> > > >   mzero = ((vector float)vec_sl(vu32(mzero), vu32(mzero)));
> > > >
> > > > got to be more happening.  If not I have made the changes and it at
> > > > least compiles now.
> > >
> > >  No, that is exactly what is happening.  Please try sending your patch.
> >
> > See attached patch.
> 
> ok, can someone with ppc confirm that this still works? (same md5 when 
> encoding something to mpeg, NOT regression tests)

I'd love to test this, but I don't have the patch around anymore.  If
somebody could send it to me I'd love to give it a spin.  An example
command line for testing would also help.

Diego