[MPlayer-cvslog] r24309 - in trunk/libswscale: swscale.c swscale_internal.h yuv2rgb.c

[Reimar Döffinger]

Hello,
On Fri, Aug 31, 2007 at 03:43:45PM -0400, Marc Hoffman wrote:
> On 8/31/07, Rich Felker <dalias at aerifal.cx> wrote:
> > On Fri, Aug 31, 2007 at 04:57:50PM +0200, mhoffman wrote:
> > > Author: mhoffman
> > > Date: Fri Aug 31 16:57:50 2007
> > > New Revision: 24309
> > >
> > > Log:
> > > renaming ARCH_BFIN to HAVE_BFIN
> >
> > Is bfin an arch or a feature of another cpu type? This change seems
> > suspicious..
> 
> I guess this must have confused me, what is the correct way for me to
> solve this configuration issue?
> 
> #ifdef HAVE_MMX2
>     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
> #elif defined (HAVE_3DNOW)
>     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
> #elif defined (HAVE_MMX)
>     flags |= SWS_CPU_CAPS_MMX;
> #elif defined (HAVE_ALTIVEC)
>     flags |= SWS_CPU_CAPS_ALTIVEC;
> #elif defined (HAVE_BFIN)
>     flags |= SWS_CPU_CAPS_BFIN;
> #endif
> #endif /* RUNTIME_CPUDETECT */

Hmm... what the heck is this doing in runtime cpu detection?
Interestingly BFIN is in _cpuexts_all, but it is never set in
_cpuexts...
I do not know what patch you use to add bfin support, but I'd think that
you should add BFIN to _arch_all and set _arch to BFIN (and remove it
from _cpuexts_all/_cpuexts).

Greetings,
Reimar Döffinger