[MPlayer-dev-eng] Altivec+GCC fix

[Jens Axboe]

On Fri, May 23 2003, Magnus Damm wrote:
> Hi again,
> 
> > > > * libmpeg2/idct_altivec.c
> > > >   Updated to version 1.19 from sourceforge
> > > > * libmpeg2/motion_comp_altivec.c
> > > >   Updated to version 1.12 from sourceforge
> > 
> > However, these new files require altivec flags to be present to be able
> > to compile. That is why I modified the Makefile instead of throwing in
> > ifdefs in the source files. I though that it was easier to just have
> > one file (makefile) that differs instead of modifying each c-file.
> > 
> > To verify this I will download and test the libmpeg2 cvs
> > on sourceforge, but I have no time today - maybe tomorrow.
> 
> After downloading the libmpeg2 snapshot and building it I realized that
> libmpeg2-configure checks for altivec.h and enables altivec if that file
> is present. And it is not possible to pass --enable/disable-altivec to 
> force support on/off.
> 
> I can't see any makefiles/conditions that excludes idct_altivec.c and 
> motion_comp_altivec.c from the build process if altivec support is missing,
> and it is not possible for me to build these altivec-enabled files without 
> passing the altivec-flags to gcc.
> 
> My conclusion is that it should be impossible to build libmpeg2 (on sf)
> for powerpc with a gcc that has no support for altivec, but I have not
> it since I have no compiler without altivec support.
> 
> mplayer has the nice feature --disable/enable-altivec so we need to 
> conditionally choose the altivec sources depending on if altivec is 
> enabled or disabled.
> 
> So I would say my patch is correct... =)

I played some with this patch yesterday, and it makes a world of
difference. DVD playback is basically not usable in current mplayer cvs
without it.

Just consider it a plea for adding the patch. Tested perfectly here.
Compiler used was 3.3 (-release).

-- 
Jens Axboe